Fact-checked by Grok 2 weeks ago

Trajectory optimization

Trajectory optimization is the process of determining a sequence of states and controls for a dynamic system that minimizes or maximizes a performance index, subject to dynamic constraints, conditions, and possibly path constraints. The performance index is typically an cost function measuring aspects such as time, , or distance traveled, while dynamic constraints are expressed as equations governing the system's evolution over time, with conditions specifying initial and final states. Path constraints further restrict states and controls along the trajectory to ensure feasibility, such as input limits or obstacle avoidance. As a core subfield of theory, trajectory optimization addresses finite-horizon problems where the goal is to compute an open-loop or from a specific , often parameterized over time rather than the entire state space. Problems are generally formulated in continuous time but solved numerically via , leading to large-scale nonlinear programs. Key solution approaches include indirect methods, which use necessary conditions from the like to derive boundary-value problems, and direct methods, which transcribe the infinite-dimensional problem into a finite-dimensional optimization via techniques such as , , or pseudospectral approximation. Direct , for instance, approximates states and controls with piecewise polynomials and enforces dynamics at points for computational efficiency. Trajectory optimization finds widespread application in fields requiring precise under constraints, including for , planetary landing, and orbit transfers; for manipulator motion and ; and autonomous vehicles for path planning. In space missions, it enables fuel-efficient powered descent guidance while respecting thrust bounds and glideslope constraints. In , it generates smooth trajectories for tasks like swing-up maneuvers or obstacle navigation, often integrated with for real-time adaptation. Advances in and successive convexification have made these methods suitable for implementation in complex, nonconvex environments.

Fundamentals

Definition and Problem Statement

Trajectory optimization is a subfield of theory focused on determining the optimal sequence of states and controls for a dynamic system to minimize a , while satisfying the system's governing dynamics and a variety of constraints. This approach computes open-loop solutions that define the best path or over a finite , often for systems where computing closed-loop policies is computationally intensive. The general problem statement involves minimizing a performance index, or cost functional, subject to dynamic constraints and boundary conditions. Mathematically, this entails finding states \mathbf{x}(t) and controls \mathbf{u}(t) that minimize J[\mathbf{x}, \mathbf{u}] = \phi(\mathbf{x}(t_f)) + \int_{t_0}^{t_f} L(\mathbf{x}(t), \mathbf{u}(t), t) \, dt, where \phi is the terminal , L is the running (), t_0 and t_f are the initial and final times, subject to the dynamics \dot{\mathbf{x}}(t) = \mathbf{f}(\mathbf{x}(t), \mathbf{u}(t), t), initial and terminal boundary conditions (e.g., \mathbf{x}(t_0) = \mathbf{x}_0, \mathbf{x}(t_f) = \mathbf{x}_f), and inequality constraints such as path inequalities \mathbf{g}(\mathbf{x}(t), \mathbf{u}(t), t) \leq \mathbf{0} and bounds on states and controls. This formulation distinguishes trajectory optimization from related fields like path planning, which typically addresses geometric feasibility without explicit dynamic models or time-dependent costs, and from static optimization, which lacks the temporal evolution governed by differential equations. Common objectives include fuel minimization for ascent trajectories, time-optimal for robotic manipulators, and energy-efficient paths for autonomous vehicles.

Mathematical Formulation

The trajectory optimization problem is commonly formulated in continuous time as the Bolza problem, which seeks to minimize a functional comprising both an running and a , subject to dynamic constraints and conditions. Specifically, the objective is to find the trajectory \mathbf{x}(t) \in \mathbb{R}^n and trajectory \mathbf{u}(t) \in \mathbb{R}^m over the time [t_0, t_f] that minimize J = \int_{t_0}^{t_f} L(t, \mathbf{x}(t), \mathbf{u}(t)) \, dt + \Phi(t_f, \mathbf{x}(t_f)), where L: \mathbb{R} \times \mathbb{R}^n \times \mathbb{R}^m \to \mathbb{R} is the representing the running (e.g., effort or ), and \Phi: \mathbb{R} \times \mathbb{R}^n \to \mathbb{R} is the function. This minimization is subject to the system dynamics \dot{\mathbf{x}}(t) = \mathbf{f}(t, \mathbf{x}(t), \mathbf{u}(t)), \quad \mathbf{x}(t_0) = \mathbf{x}_0, where \mathbf{f}: \mathbb{R} \times \mathbb{R}^n \times \mathbb{R}^m \to \mathbb{R}^n describes the evolution of the variables, typically derived from physical laws such as Newton's equations in or in . Additional constraints include conditions \psi(t_0, \mathbf{x}(t_0)) = \mathbf{0}, conditions \phi(t_f, \mathbf{x}(t_f)) = \mathbf{0}, and constraints \mathbf{g}(t, \mathbf{x}(t), \mathbf{u}(t)) \leq \mathbf{0} to enforce feasibility, such as or bounds. Here, the state \mathbf{x}(t) captures the system's configuration and velocities (e.g., position and velocity in a double integrator model), while the control \mathbf{u}(t) represents actionable inputs (e.g., thrust or torque). The initial time t_0 is typically fixed, but the final time t_f may be free, requiring an additional transversality condition in the optimization. Control constraints are often incorporated via \mathbf{u}(t) \in \mathcal{U}, a compact set ensuring physical realizability. Variations of this formulation adapt to specific problem structures. The Mayer form omits the integral cost by setting L \equiv 0, focusing solely on \min \Phi(t_f, \mathbf{x}(t_f)), which simplifies certain numerical implementations. Conversely, the Lagrange form sets \Phi \equiv 0, emphasizing the running cost \min \int_{t_0}^{t_f} L \, dt. Isoperimetric constraints introduce auxiliary integral conditions, such as \int_{t_0}^{t_f} \mathbf{h}(t, \mathbf{x}(t), \mathbf{u}(t)) \, dt = \mathbf{c}, to enforce conserved quantities like total energy. For computational solution, the continuous problem is often discretized into a nonlinear programming (NLP) form using time-stepping methods, such as the explicit Euler scheme where states evolve as \mathbf{x}_{k+1} = \mathbf{x}_k + h \mathbf{f}(t_k, \mathbf{x}_k, \mathbf{u}_k) for discrete times t_k = t_0 + k h and step size h, transforming the trajectories into finite-dimensional decision variables subject to algebraic constraints.

Historical Development

Early Foundations

The foundations of trajectory optimization trace back to the , a branch of mathematics developed in the to find curves or paths that extremize certain functionals. Leonhard Euler advanced this field in 1736 by addressing the brachistochrone problem, which seeks the curve of fastest descent between two points under gravity, demonstrating its solution as a and laying groundwork for variational methods applicable to path optimization. Building on Euler's ideas, Joseph-Louis formalized the Euler-Lagrange equations in his 1788 treatise Mécanique Analytique, providing a systematic framework for deriving that minimize or maximize integrals along static paths, essential for early trajectory problems in mechanics. A key transition to dynamic systems occurred with William Rowan Hamilton's principle, introduced in 1834, which reformulates variational mechanics by stating that the actual path of a system extremizes the action integral defined by the , bridging static variational calculus to time-dependent trajectories in . This principle influenced subsequent developments, including precursors to the Hamilton-Jacobi equation in the mid-19th century, where Hamilton's 1834-1835 essays and Carl Gustav Jacob Jacobi's 1837 generalizations provided formulations for optimal paths in conservative systems, foreshadowing later dynamic programming approaches. further extended the theory in the 1870s through his lectures on , introducing transversality conditions that ensure optimality at boundaries for variable endpoint problems, refining the handling of constraints in path optimization. Early engineering applications emerged in rocketry, exemplified by Robert H. 's 1919 calculations for multi-stage rocket trajectories aimed at extreme altitudes. In his seminal paper, Goddard derived fuel-optimal ascent paths using approximate solutions to differential equations for , , and , incorporating step-wise to maximize velocity and range, marking one of the first practical applications of optimization to dynamic rocket trajectories.

Key Milestones in Optimal Control

The development of dynamic programming in the 1950s provided a foundational framework for solving trajectory optimization problems through recursive decomposition. Richard Bellman introduced of optimality, which states that an optimal has the that, regardless of the initial state and decision, the remaining decisions must constitute an optimal for the resulting state. This principle underpins the value function V(x,t), defined as the minimum cost-to-go from state x at time t, satisfying V(x,t) = \min_u \left[ \int_t^{t_f} L(x,\dot{x},u) \, dt + V(x_f, t_f) \right], where L is the running cost and the integral is over the trajectory. Bellman's work, detailed in his 1957 book Dynamic Programming, enabled the computational solution of multistage decision problems in control systems, marking a shift toward discrete-time approximations for continuous trajectory problems. In 1956, and his collaborators formulated the , establishing necessary conditions for optimality in continuous-time control problems central to trajectory optimization. The principle involves the H(x, u, \lambda, t) = \lambda^T f(x, u, t) - L(x, u, t), where f describes the \dot{x} = f(x, u, t), and the costate \lambda evolves as \dot{\lambda} = -\frac{\partial H}{\partial x}. The optimal control u^* maximizes H over the admissible set at each instant. This analytic tool, first presented in Pontryagin's 1961 monograph The Mathematical Theory of Optimal Processes (reflecting 1956 origins), provided a rigorous basis for indirect methods, influencing subsequent derivations of optimality conditions without requiring full trajectory parameterization. The 1960s accelerated practical applications of to trajectory optimization, particularly for NASA's Apollo missions. Engineers applied Pontryagin's principle and dynamic programming to compute fuel-efficient lunar transfers and reentry trajectories, ensuring precise guidance under constraints like thrust limits and atmospheric heating. Arthur E. Bryson and Yu-Chi Ho's 1969 textbook Applied Optimal Control: Optimization, Estimation, and Control synthesized these advancements, offering algorithms for two-point boundary value problems encountered in , including iterative shooting methods for Apollo-like transfers. The book emphasized computational implementation, bridging theory and engineering practice during an era of rapid mission demands. Numerical breakthroughs in the 1970s facilitated the rise of direct methods for trajectory optimization, complementing indirect approaches by discretizing problems into nonlinear programs solvable via gradient-based solvers. Donald E. Kirk's 1970 textbook Optimal Control Theory: An Introduction detailed these techniques, covering for state trajectories and for cost functionals, with examples in . This period saw increased accessibility due to advancing computers, enabling direct collocation schemes that approximated controls and states on finite grids, reducing sensitivity to initial guesses compared to earlier methods. Pseudospectral methods emerged in the as a high-accuracy direct approach for trajectory optimization, leveraging global approximations like Chebyshev or Legendre basis functions to achieve convergence. These methods discretize the entire time domain at points, transforming differential equations into algebraic constraints for efficient . Early developments, including Legendre pseudospectral formulations, demonstrated superior performance for constrained problems, such as reentry trajectories, by minimizing mesh dependencies. In the , open-source tools democratized trajectory optimization, with the ACADO Toolkit providing a C++ framework for automatic code generation in optimal control. Released around 2010, ACADO supports multiple shooting, , and real-time methods for nonlinear problems, including for dynamic trajectories. Its integration with solvers like qpOASES enabled widespread adoption in and automotive applications, fostering reproducible research and reducing barriers to implementing complex optimizations. In the 2020s, trajectory optimization has seen significant integration with for data-driven methods and applications, alongside the proliferation of advanced open-source libraries such as Crocoddyl for and OpenAP.top for trajectory planning, enhancing scalability and adaptability as of 2025.

Applications

Aerospace Applications

Trajectory optimization plays a crucial role in , particularly for missions involving and where minimizing consumption, time, or thermal loads is essential under complex dynamic constraints like , atmosphere, and thrust limits. In , it enables efficient ascent from launch pads to , precise orbital maneuvers, and safe planetary entries, while in atmospheric flight, it supports rapid climbs and energy-efficient paths for jets and unmanned aerial vehicles (UAVs). These applications often leverage indirect methods based on theory to derive bang-bang thrust profiles or continuous adjustments that balance competing objectives. Rocket ascent trajectories, exemplified by the classic Goddard's problem, focus on maximizing altitude or for a vertically launching under variable and influences, with constraints on magnitude to minimize fuel use and account for losses. In multi-stage launch vehicles like those used in modern missions, optimization sequences stage firings to achieve orbital insertion while respecting dynamic pressure limits during ascent, improving overall compared to heuristic profiles. For instance, the Goddard problem formulation treats the rocket's motion as a one-dimensional task, solving for profiles that avoid singular arcs and ensure feasibility under inverse-square . Orbital transfers rely on trajectory optimization to minimize delta-v, the change in velocity required for maneuvers, often using impulsive approximations in two-body dynamics. The Hohmann transfer represents a baseline elliptical path between circular orbits, requiring a total delta-v of approximately \sqrt{\frac{\mu}{r_1}} \left( \sqrt{\frac{2 r_2}{r_1 + r_2}} - 1 \right) + \sqrt{\frac{\mu}{r_2}} \left( 1 - \sqrt{\frac{2 r_1}{r_1 + r_2}} \right), where \mu is the gravitational parameter and r_1, r_2 are the initial and final radii; this minimizes fuel for coplanar transfers by tangent burns at periapsis and apoapsis. For scenarios, extends this by solving for the connecting two position vectors in a specified time, enabling multi-revolution solutions that reduce delta-v by 5-20% over single-revolution paths in perturbed environments. Reentry and landing trajectories demand optimization to manage hypersonic aerothermal loads while achieving precise , particularly for planetary missions with thin atmospheres like Mars. Hypersonic glide vehicles use bank-angle to steer along energy-dissipating paths, balancing peak heating rates below 100 W/cm² and deceleration forces up to about 10 to protect payloads. In the 2021 Perseverance rover , trajectory optimization via predictor-corrector guidance adjusted entry interface conditions to target Jezero Crater within a 7.7 km × 6.4 km , incorporating terrain-relative to mitigate uncertainties in wind and density, resulting in a final position error of less than 100 m. This approach integrated six-degree-of-freedom dynamics from entry to powered descent, optimizing for minimum fuel while constraining and at parachute deploy. Atmospheric flight applications emphasize time or in climb and phases, adapting to variable winds and engine performance. For , minimum-time-to-climb trajectories follow energy-state approximation paths, accelerating at constant in the subsonic regime before transitioning to constant-equivalent climbs, achieving altitudes like 11 km in under 10 minutes while minimizing . UAVs incorporate wind-aware path planning to exploit tailwinds for extended , using receding-horizon optimization to generate feasible routes that reduce expenditure by 20-30% in gusty conditions, ensuring collision avoidance and battery constraints. A notable is the translunar injection (TLI) maneuver, optimized using to transition from Earth to a lunar with minimal delta-v of about 3.1 km/s. Reconstruction of the 1969 mission via modern numerical optimization confirms the original indirect method's efficacy, solving the two-point boundary-value problem for thrust steering that accounted for spherical gravity and third-body perturbations, achieving insertion within 0.1% of targeted perilune altitude. This application highlighted the principle's ability to handle free-final-time problems, influencing subsequent lunar mission designs.

Robotics Applications

In robotic manipulators, trajectory optimization is employed to generate joint-space paths that minimize energy consumption or execution time while integrating to ensure end-effector accuracy, particularly in industrial arms like the KR 16 for pick-and-place tasks. For instance, optimization techniques have been applied to reduce energy use by formulating the problem as a nonlinear program that respects joint limits and velocity constraints, achieving up to 20% savings in power draw compared to standard trapezoidal profiles. This approach contrasts with pure kinematic planning by incorporating dynamic models to balance smoothness and efficiency. For quadrotor helicopters, trajectory optimization enables agile flight paths that incorporate obstacle avoidance and attitude control under aerodynamic constraints, as demonstrated in maneuvers inspired by DARPA's Subterranean where drones navigated cluttered underground environments. Methods such as minimum-snap trajectory generation with collision-free polynomials allow replanning, enabling quadrotors to perform agile maneuvers while avoiding obstacles. These techniques prioritize dynamic feasibility, ensuring stable hovering and rapid recovery from perturbations. In walking robots, gait optimization via trajectory planning incorporates zero-moment point (ZMP) constraints to maintain stability in bipedal locomotion, exemplified by the Boston Dynamics Atlas robot where online replanning generates smooth footstep sequences for dynamic tasks like jumping or rough-terrain traversal. Seminal work on Atlas used direct collocation to optimize center-of-mass trajectories, enforcing ZMP within the support polygon and achieving sub-millisecond computation times for real-time adaptation, which supported robust performance in simulated DARPA Robotics Challenge scenarios. This ensures energy-efficient gaits by minimizing torque variations across phases of single- and double-support. Swarm robotics leverages trajectory optimization for coordinated multi-agent systems, focusing on formation control where agents maintain relative positions while avoiding collisions, often using distributed algorithms to scale to dozens of units. For example, leader-follower frameworks optimize collective paths by solving coupled nonlinear programs that balance formation integrity and obstacle clearance, as in quadrotor swarms where trajectories are parameterized by B-splines to achieve collision-free flights at densities up to 10 agents per cubic meter. These methods enhance robustness through decentralized computation, reducing global communication overhead. A notable case study is the HRP-2 humanoid robot, where direct collocation has been used for trajectory planning in walking motions, transcribing the optimal control problem into a nonlinear program that optimizes joint torques and contact forces while respecting ZMP stability. Applied to HRP-2, this approach generated efficient gaits for multi-contact scenarios, such as stair climbing, by discretizing dynamics over time grids and solving via , resulting in trajectories that reduced energy expenditure relative to heuristic methods. The framework's scalability allowed integration with whole-body control for real-time execution on the physical platform.

Industrial Applications

In manufacturing processes, trajectory optimization plays a crucial role in tool path planning for computer numerical control (CNC) , where the goal is to generate smooth, time-optimal paths that minimize machining time and while adhering to geometric and kinematic constraints. These optimizations often involve formulating the tool trajectory as a constrained problem, ensuring continuous velocity and profiles to reduce and extend tool life in high-speed operations. For instance, methods that parameterize trajectories using splines or polynomials have been shown to reduce cycle times compared to traditional linear interpolations, enhancing overall production efficiency in industries like automotive part fabrication. In chemical processing, trajectory optimization is applied to batch reactors, particularly in fed-batch fermentation, to determine optimal temperature and feed profiles that maximize product yield and resource utilization. By solving dynamic optimization problems, these approaches adjust temperature trajectories to balance reaction kinetics and microbial growth, often resulting in higher or pharmaceutical yields and improvements in productivity for recombinant . This is achieved through indirect methods like or direct collocation techniques, which handle nonlinear dynamics inherent in biochemical processes. The leverages trajectory optimization for autonomous vehicle driving, focusing on economy under traffic and environmental constraints, as seen in systems developed by companies like in the 2020s. These optimizations generate energy-efficient speed and lane-change trajectories, incorporating models to minimize consumption—demonstrating reductions of 5-10% in driving scenarios through predictive control that anticipates . Such applications integrate from sensors and V2X communications to ensure safe, smooth paths that align with industrial standards for fleet operations. In energy systems, trajectory optimization supports industrial operations like wind turbine blade inspections using drones, where paths are planned to minimize energy use while covering critical surfaces under wind disturbances. Optimized drone trajectories, often computed via mixed-integer programming, enable comprehensive scans with reduced flight times and battery consumption in offshore farm inspections by avoiding turbulent zones. This enhances maintenance efficiency and operational uptime in renewable energy production. A notable is wafer processing, where trajectory optimization designs heating and cooling cycles for rapid thermal annealing () to achieve uniform temperature distributions and minimize defects. strategies determine lamp power trajectories that ramp temperatures precisely, reducing thermal gradients across the wafer and improving quality for ultrashallow implants essential in advanced fabrication. These methods prioritize minimal overshoot and to boost throughput in high-volume .

Key Concepts and Terminology

Core Definitions

In trajectory optimization, the state refers to the sequence of states x(t) that describes the evolution of a over time, typically from an to a terminal state within a specified horizon. This captures the system's and velocities as it moves through its state space, governed by the underlying . The trajectory, denoted as u(t), consists of the time-varying inputs applied to the to steer it along the desired state trajectory, often parameterized over the same time interval. These inputs represent actuators or forces that influence the 's behavior while respecting operational limits. The cost functional, commonly expressed as J, quantifies the overall performance of a trajectory by integrating or summing a measure of deviation from desired behavior, such as terms that penalize excessive effort or deviations from . It serves as the objective to be minimized, balancing trade-offs like energy use and path efficiency. A feasible is any and pair that satisfies the system's and all imposed constraints, such as bounds on states, controls, or intermediate conditions, without regard to performance optimization. Feasibility ensures the trajectory is physically realizable within the problem's boundaries. Optimality describes a trajectory where the cost functional achieves a minimum value, such that no other feasible trajectory yields a lower , either locally (no small perturbations improve it) or globally (no better trajectory exists overall). This condition defines the solution to the trajectory optimization problem.

Constraints and Objectives

In trajectory optimization, constraints define the feasible set of state and control trajectories that must satisfy physical, operational, or requirements, while objectives specify the performance criteria to minimize or maximize along those trajectories. Path constraints typically take the form of time-varying inequalities, such as g(x(t), u(t), t) \leq 0, which enforce limits like maximum or heating rates during flight to ensure without violating system capabilities. These constraints introduce by coupling the x(t) and u(t) over the entire , often requiring careful parameterization to maintain computational tractability. Control bounds represent simple box constraints on the control inputs, such as u_{\min} \leq u(t) \leq u_{\max}, which model actuator saturation limits like thrust bounds in rocket propulsion or torque limits in robotic manipulators. These are ubiquitous in practical problems, as exceeding them can lead to system failure, and they are typically handled directly in optimization formulations to prevent infeasible solutions. State constraints, often expressed as h(x(t)) \leq 0, pose greater challenges due to their potential non-convexity, as seen in obstacle avoidance where the feasible region excludes collision zones, necessitating specialized techniques like sequential convexification to approximate and resolve the non-convex sets. Objectives in trajectory optimization are formulated as cost functions J = \phi(x(T)) + \int_{t_0}^{T} L(x(t), u(t), t) \, dt, where \phi is the terminal cost and L is the running cost, aiming to balance trade-offs like versus time. Constraints can be classified as hard, which must be strictly satisfied (e.g., bounds), or soft, which are penalized within the objective via Lagrange multipliers or slack variables to allow minor violations for overall optimality. In multi-objective settings, conflicting goals such as minimizing energy and time are resolved through Pareto optimization, generating a set of non-dominated solutions that represent trade-offs without a single optimum. Terminal constraints specify conditions at the final time T, such as fixed endpoints x(T) = x_f for precise in , contrasting with free endpoints where transversality conditions from the Pontryagin dictate that the costate p(T) aligns with the of the terminal set to ensure optimality. These fixed terminal constraints enforce mission-specific goals, like achieving a target in applications, while free cases allow flexibility at the expense of additional boundary conditions in the solution process.

Optimization Methods

Indirect Methods

Indirect methods for trajectory optimization derive analytical necessary conditions for optimality from the and theory, transforming the continuous-time optimal control problem into a two-point boundary value problem (TPBVP) defined by coupled state and costate differential equations. These conditions, when satisfied, ensure that the solution is a of the objective functional, often with guarantees of local optimality. The approach emphasizes theoretical rigor over direct numerical approximation, requiring subsequent numerical solution of the TPBVP to obtain the explicit trajectory. The roots of indirect methods trace to the calculus of variations, which provides necessary conditions for unconstrained trajectory problems minimizing a cost functional of the form J[\mathbf{x}] = \int_{t_0}^{t_f} L(\mathbf{x}(t), \dot{\mathbf{x}}(t), t) \, dt, with fixed boundary conditions \mathbf{x}(t_0) = \mathbf{x}_0 and \mathbf{x}(t_f) = \mathbf{x}_f. To derive the optimality conditions, consider a variation \delta \mathbf{x}(t) around a candidate trajectory \mathbf{x}(t), leading to the first variation \delta J = \left[ \frac{\partial L}{\partial \dot{\mathbf{x}}} \delta \mathbf{x} \right]_{t_0}^{t_f} + \int_{t_0}^{t_f} \left( \frac{\partial L}{\partial \mathbf{x}} - \frac{d}{dt} \frac{\partial L}{\partial \dot{\mathbf{x}}} \right) \delta \mathbf{x}(t) \, dt = 0. For arbitrary admissible variations vanishing at the boundaries, the integrand must be zero, yielding the Euler-Lagrange equations \frac{d}{dt} \left( \frac{\partial L}{\partial \dot{\mathbf{x}}} \right) = \frac{\partial L}{\partial \mathbf{x}}. These second-order differential equations describe the necessary dynamics for an extremal trajectory in the state space, solvable as a TPBVP for fixed endpoints. For controlled systems, where dynamics are governed by \dot{\mathbf{x}} = \mathbf{f}(\mathbf{x}, \mathbf{u}, t) with \mathbf{u}(t) \in U (a compact set) and the is J = \phi(\mathbf{x}(t_f)) + \int_{t_0}^{t_f} L(\mathbf{x}, \mathbf{u}, t) \, dt, (PMP) generalizes the Euler-Lagrange conditions to include explicit . The augmented is H(\mathbf{x}, \mathbf{u}, \boldsymbol{\lambda}, \nu, t) = \nu L(\mathbf{x}, \mathbf{u}, t) + \boldsymbol{\lambda}^T \mathbf{f}(\mathbf{x}, \mathbf{u}, t), where \boldsymbol{\lambda}(t) \in \mathbb{R}^n are costate variables and \nu \geq 0 is a constant multiplier with (\nu, \boldsymbol{\lambda}(t)) \neq \mathbf{0}. The PMP asserts that along an optimal trajectory (\mathbf{x}^*, \mathbf{u}^*), the following hold for almost all t \in [t_0, t_f]:
  • State dynamics: \dot{\mathbf{x}}^* = \frac{\partial H}{\partial \boldsymbol{\lambda}} \big|_{\mathbf{x}^*, \mathbf{u}^*, \boldsymbol{\lambda}^*, \nu, t},
  • Costate dynamics (adjoint equations): \dot{\boldsymbol{\lambda}}^* = -\frac{\partial H}{\partial \mathbf{x}} \big|_{\mathbf{x}^*, \mathbf{u}^*, \boldsymbol{\lambda}^*, \nu, t},
  • Control optimality: \mathbf{u}^*(t) = \arg\max_{\mathbf{u} \in U} H(\mathbf{x}^*(t), \mathbf{u}, \boldsymbol{\lambda}^*(t), \nu, t),
along with transversality conditions at t_f, such as \boldsymbol{\lambda}^*(t_f) = \nu \frac{\partial \phi}{\partial \mathbf{x}} \big|_{\mathbf{x}^*(t_f)} for fixed t_f and free terminal state (or adjusted for other cases). The multiplier \nu = 0 corresponds to pure state maximization problems (e.g., time-optimal), while \nu = 1 normalizes typical Mayer-Bolza forms; nontriviality ensures the conditions are not degenerate. This principle is derived by augmenting the cost with the dynamic constraints via Lagrange multipliers \boldsymbol{\lambda}(t), forming the variational problem \tilde{J} = \phi + \int (\nu L + \boldsymbol{\lambda}^T (\mathbf{f} - \dot{\mathbf{x}})) dt. Setting \delta \tilde{J} = 0 for variations in state, control, and multipliers yields the stationarity conditions. In continuous time, the derivation proceeds via on the state variation term, producing the , while the control variation requires the to be maximized to achieve a minimum (or maximum, per ) over U. For convex U, the maximizer often yields explicit bang-bang or singular controls; the full continuous-time form emerges rigorously as the limit of discrete-time approximations where time steps \Delta t \to 0. The resulting TPBVP from PMP—comprising $2nfirst-order ODEs for(\mathbf{x}, \boldsymbol{\lambda})withninitial conditions known andnterminal conditions to satisfy—is typically solved using indirect [shooting](/page/Shooting). An initial guess\boldsymbol{\lambda}(t_0)is selected (along with fixed\mathbf{x}(t_0)), and the coupled [system](/page/System) is integrated forward from t_0tot_fusing a stiff [ODE](/page/Ode) solver (e.g., implicit Runge-Kutta). The terminal mismatch\boldsymbol{\psi}(\boldsymbol{\lambda}(t_0)) = \mathbf{g}(\mathbf{x}(t_f), \boldsymbol{\lambda}(t_f)) = \mathbf{0}(encoding transversality and endpoint constraints) is computed, and\boldsymbol{\lambda}(t_0)is updated iteratively via root-finding, such as simple [shooting](/page/Shooting) (gradient descent on|\boldsymbol{\psi}|^2$) or multiple (segmented integration with continuity constraints). via variational equations aids convergence in higher dimensions. Indirect methods provide strong theoretical advantages, including certification that solutions satisfy optimality conditions exactly (up to numerical tolerance), and global optimality guarantees when the is in and the problem is overall. However, they are disadvantaged by extreme sensitivity to initial costate guesses, often requiring domain expertise or continuation to converge, especially in high-dimensional or non- settings where multiple local extrema exist. A canonical example is time-optimal control for linear systems \dot{\mathbf{x}} = A \mathbf{x} + B \mathbf{u}, |\mathbf{u}| \leq 1, minimizing t_f subject to \mathbf{x}(t_0) = \mathbf{x}_0, \mathbf{x}(t_f) = \mathbf{0}. PMP yields \nu = 0, H = \boldsymbol{\lambda}^T (A \mathbf{x} + B \mathbf{u}), adjoint \dot{\boldsymbol{\lambda}} = -A^T \boldsymbol{\lambda}, and bang-bang control \mathbf{u}^* = \operatorname{sign}(B^T e^{A^T (t_f - t)} \boldsymbol{\lambda}(t_f)). For the double integrator (\ddot{y} = u, |u| \leq 1, from (y(0), \dot{y}(0)) = (0, 0) to (y(t_f), \dot{y}(t_f)) = (y_f, 0)), the optimal policy switches once, producing a parabolic position trajectory y(t) = \frac{1}{2} t^2 for $0 \leq t \leq t_s followed by deceleration, achieving y_f = \frac{1}{4} t_f^2 in minimum time t_f = 2 \sqrt{y_f}. This illustrates singular arcs avoidance and explicit switch times from costate geometry.

Direct Methods

Direct methods for trajectory optimization discretize the continuous-time problem into a finite-dimensional (NLP) problem, which is then solved using general-purpose numerical solvers. This transcription approach parameterizes the state \mathbf{x}(t) and control \mathbf{u}(t) at time nodes t_k for k = 0, \dots, N, forming a decision \mathbf{z} = [\mathbf{x}_0, \mathbf{u}_0, \dots, \mathbf{x}_N, \mathbf{u}_N]^\top. The objective is reformulated as minimizing a cost J_d(\mathbf{z}), typically a approximation of the cost, subject to discretized constraints \mathbf{g}(\mathbf{z}) = \mathbf{0} and bounds on \mathbf{z}. This enables handling complex nonlinear and constraints without deriving analytical necessary conditions. A key variant is the multiple shooting method, which segments the time horizon into multiple intervals and solves boundary value problems by treating initial states of each segment as optimization variables. Continuity of states is enforced at junction nodes between segments, improving stability and allowing parallel computation compared to single shooting. Mesh refinement techniques adaptively adjust node density based on error estimates, enhancing efficiency for problems with varying solution smoothness. These features make multiple shooting particularly effective for long-duration trajectories in aerospace applications. Collocation methods approximate the trajectories with local basis functions, such as polynomials over short , and enforce the exactly at points within each . The offers a low-order by assuming linear variation in the : \dot{\mathbf{x}}_k \approx \frac{\mathbf{x}_{k+1} - \mathbf{x}_k}{h} = \frac{f(\mathbf{x}_k, \mathbf{u}_k) + f(\mathbf{x}_{k+1}, \mathbf{u}_{k+1})}{2}, where h is the step size and f is the system ; this leads to simple algebraic constraints in the . For higher accuracy, the Hermite-Simpson method employs cubic Hermite splines for states, collocating the defect (dynamics residual) at endpoints and the , resulting in continuous first derivatives and better properties for mechanical systems. These methods balance computational cost and precision, with Hermite-Simpson often preferred for its quadratic convergence. Pseudospectral methods provide global approximations using high-order orthogonal polynomials over the entire domain, achieving spectral (exponential) convergence with modest node counts. occurs at Legendre-Gauss-Radau (LGR) nodes, which are roots of the derivative of the Legendre polynomial shifted to include the final boundary. Time derivatives are computed via differentiation matrices \mathbf{D}, such that \dot{\mathbf{x}}(\tau_i) \approx \mathbf{D} \mathbf{x}, where \tau_i are normalized points. The are enforced as \mathbf{D} \mathbf{x} - f(\mathbf{x}, \mathbf{u}, \tau) = \mathbf{0} at interior nodes, with boundary conditions handled separately. LGR formulations excel in problems requiring high accuracy, such as reentry trajectories, due to their ability to capture sharp gradients efficiently. Additional direct techniques encompass temporal finite element methods, which discretize the weak form of the using finite elements in time to yield variational integrators that preserve symplectic structure and energy-like quantities in systems. (DDP) iteratively linearizes the dynamics and quadratizes the cost around a nominal , solving local value functions backward and refining controls forward for second-order convergence in smooth problems. In the , diffusion-based methods have emerged, employing score matching to sample trajectories from a learned conditioned on constraints, integrating to handle high-dimensional or uncertain environments. A representative example is the double integrator \ddot{q} = u with |u| \leq [1](/page/1), minimizing time to reach a target from rest. Direct discretizes q_k and v_k at nodes, approximating via the and enforcing v_{k+1} - v_k = h u_k and q_{k+1} - q_k = h (v_k + v_{k+1})/2 under , yielding an NLP whose solution reveals a bang-bang profile with one switch. This setup demonstrates how transforms the problem into a solvable form, scalable to more complex systems like maneuvers.

Comparison of Methods

Indirect versus Direct Approaches

Indirect methods in trajectory optimization derive from the necessary conditions of optimality provided by theory, such as , which yield exact conditions for the optimal solution in the form of a two-point (BVP). These methods provide a theoretical certification of optimality when a solution is found, as they satisfy the first-order necessary conditions analytically before numerical solution. In contrast, direct methods approximate the continuous optimal control problem by discretizing the state and control variables, transforming it into a finite-dimensional (NLP) problem that inherently includes discretization errors, lacking the same level of theoretical exactness. Computationally, indirect methods involve solving the BVP through techniques like multiple shooting, often using integrators such as Runge-Kutta to propagate the state and adjoint equations while enforcing boundary conditions. Direct methods, however, transcribe the problem into an and employ specialized solvers like or SNOPT to optimize the discretized variables, making them more straightforward to implement for complex formulations. Indirect approaches typically require careful initialization of costates, whereas direct methods benefit from broader applicability to problems with inequalities and non-smooth constraints due to the flexibility of NLP frameworks. Regarding convergence, indirect methods exhibit rapid local convergence near the optimal solution but are highly sensitive to initial guesses and nonlinearities in the dynamics, often struggling with large-scale or ill-conditioned problems. Direct methods offer larger basins of attraction and better handling of non-smooth elements, though they may converge to local optima and require mesh refinement for accuracy, potentially missing global solutions in multimodal landscapes. To mitigate these limitations, hybrid approaches combine the strengths of both: direct methods generate feasible initial trajectories to initialize indirect solvers, improving convergence reliability and accuracy, as demonstrated in applications like low-thrust transfers where direct collocation seeds indirect BVPs. In terms of performance metrics, indirect methods excel in accuracy for low-dimensional problems (e.g., fewer ), achieving near-exact solutions with minimal computational overhead once converged, while direct methods scale better to high-fidelity, large-scale problems (e.g., hundreds of switches in heliocentric trajectories) but introduce approximation errors that grow with coarser discretizations. For instance, in minimum-fuel trajectory problems, nonregularized indirect formulations can yield noticeable improvements in final mass over direct methods, highlighting their edge in precision at the cost of increased sensitivity.

Discretization Strategies

In direct methods for trajectory optimization, discretization strategies transform continuous-time problems into finite-dimensional nonlinear programs by approximating states and controls. Two primary approaches are and methods. methods propagate states and controls sequentially over time intervals using , such as Runge-Kutta schemes, which can lead to accumulation of local truncation errors, particularly in long-horizon or stiff problems. In contrast, methods enforce the constraints at multiple points within each interval, typically using approximations, resulting in accuracy across the trajectory without sequential error buildup. This makes suitable for high-precision applications, while excels in scenarios requiring fewer decision variables for faster computation. Mesh strategies play a crucial role in balancing accuracy and computational cost during . Uniform divide the into equally spaced , simplifying implementation but potentially requiring many nodes for problems with varying . Adaptive refinement, however, adjusts sizes based on local estimates, concentrating nodes in regions of rapid change to improve efficiency. Advanced h-p methods combine h-refinement (increasing the number of ) with p-refinement (raising polynomial order within ), enabling automatic and for smooth solutions. These strategies are essential for handling multiphase problems or nonlinear without excessive density. Orthogonal collocation methods leverage roots of orthogonal polynomials for collocation points to achieve high-order accuracy. Radau points, derived from Legendre-Gauss-Radau quadrature, include one endpoint (typically the initial boundary) and interior points, facilitating precise boundary condition enforcement and avoiding singularities in finite- or infinite-horizon problems. This asymmetric distribution enhances stability for trajectory optimization, with spectral convergence rates where errors decay exponentially with polynomial degree. Pseudospectral methods, often using Chebyshev nodes (extrema of Chebyshev polynomials), promote faster convergence for smooth functions due to minimax properties, though they may require more points for boundary inclusion compared to Radau schemes. The choice depends on problem structure, with Chebyshev offering computational simplicity via barycentric interpolation. Finite element methods employ local piecewise bases over subintervals, yielding sparse Jacobians that exploit efficient solvers and improve numerical for large-scale problems. Pseudospectral methods, using global bases across the entire , achieve superior accuracy through but result in denser matrices, potentially worsening and increasing costs. These trade-offs make finite elements preferable for systems with discontinuities or when sparsity is critical, while pseudospectral approaches suit smooth, high-fidelity optimizations despite higher memory demands. Performance characteristics highlight distinct applications: shooting methods, with fewer optimization variables, enable real-time computation for online control, as seen in embedded systems. Collocation methods, particularly pseudospectral variants, provide higher precision for offline planning; for instance, applications demonstrated significant efficiency gains, such as computing propellant-saving maneuvers in hours on standard hardware, outperforming traditional methods in convergence speed for complex space trajectories.

Challenges and Advances

Computational Challenges

Trajectory optimization problems often suffer from the curse of dimensionality, where the grows exponentially with the number of and variables, particularly in high-degree-of-freedom (DOF) systems exceeding 10 states. This arises because methods like dynamic programming require discretizing the space, leading to an explosion in the number of grid points needed for adequate resolution; for instance, placing 10 points per dimension results in $10^n optimizations for n dimensions, rendering exact solutions infeasible for complex robotic or systems. In practice, this limits the applicability of full-order models for bipedal robots with over 10 states (e.g., angles and velocities), necessitating techniques to maintain computational tractability without sacrificing . Non-convexity poses another significant hurdle, as the resulting nonlinear programs frequently exhibit multiple local minima, trapping gradient-based solvers and complicating global optimality guarantees. This non-convexity stems from nonlinear dynamics, path constraints, or discontinuities in the objective, such as in problems with variable final times or hybrid modes, where initial guesses heavily influence convergence. via the is crucial here, as negative eigenvalues indicate points or directions of , but computing the full for large-scale problems is prohibitive due to its quadratic scaling with variables; instead, approximations like quasi-Newton methods are used, though they risk amplifying errors in non-positive-definite regions. Real-time applications, such as (MPC) in autonomous vehicles or , impose stringent timing constraints, often requiring solutions in under 100 ms per iteration to enable loops without . methods, while flexible, can demand hundreds of iterations for , exceeding these limits; approximation strategies like warm-starting—initializing the solver with the previous timestep's —reduce this by providing feasible starting points, cutting Newton steps from dozens to a handful in frameworks. However, even with warm-starting, stiff constraints or poor conditioning can still violate budgets, highlighting the need for tailored preconditioners in time-critical scenarios. Handling uncertainty further exacerbates computational demands, as robust optimization must account for parametric noise or disturbances, transforming deterministic problems into stochastic ones with chance constraints that ensure violation probabilities remain below a threshold (e.g., 5%). Stochastic variants, such as those using particle filters or sigma-point approximations, propagate uncertainty through the dynamics, but this introduces sampling overhead, scaling poorly with horizon length and uncertainty dimensions; for example, chance-constrained formulations reformulate as convex approximations yet require iterative sampling for non-Gaussian noise, increasing solve times by orders of magnitude. Robust counterparts, like worst-case min-max problems, mitigate this but often yield overly conservative trajectories, balancing feasibility against performance in noisy environments. A notable case arises in large-scale simulations, such as planetary entry guidance, where high-fidelity models with dozens of states (e.g., coupled and ) frequently fail to converge without regularization terms like penalties or trust-region bounds. These simulations, involving thousands of points over long horizons, exhibit ill-conditioning from sparse Jacobians and nonlinear constraints; regularization stabilizes the approximations, enabling , though at the cost of slight suboptimality. This underscores how indirect methods, reliant on accurate two-point boundary value solutions, amplify such issues in distributed large-scale formations like swarms.

Integration with Emerging Technologies

Trajectory optimization has increasingly integrated with to overcome challenges in initialization and dynamics modeling. Deep reinforcement learning methods, such as Deep Deterministic Policy Gradient (DDPG), provide effective initial guesses for trajectory optimization by learning policies in continuous action spaces, as applied to low-thrust trajectories and UAV path planning in the and . Neural ordinary differential equations (Neural ODEs) approximate continuous-time dynamics for more flexible modeling in optimization, enabling reduced final errors by 99% in single orbital transfer trajectories through learned guidance networks. Diffusion models facilitate generative sampling of trajectories, particularly for non-convex optimization landscapes. Approaches from 2023 align diffusion sampling trajectories with physics-based optimization steps via (DOM) and Trajectory Alignment (TA), enhancing constrained design generation by matching intermediate distributions to improve in complex path sampling. Real-time implementations leverage hardware and software for embedded systems. Field-programmable gate arrays (FPGAs) accelerate trajectory-related computations, such as parallelized for ballistic target tracking, achieving speedups of up to approximately 4-fold compared to CPU-based methods. The GPOPS-II software solves multiple-phase problems using hp-adaptive , as applied in simulations of landings. In autonomous systems, trajectory optimization underpins for self-driving cars, employing direct methods to generate collision-free paths while balancing safety and efficiency. Recent 2025 implementations optimize vehicle trajectories at intersections in , reducing delays through cooperative with signal phases. Future directions explore quantum-inspired solvers to handle large-scale trajectory problems, such as UAV trajectory optimization in low-altitude networks, by approximating .

References

  1. [1]
    [PDF] HOW TO DO YOUR OWN DIRECT COLLOCATION - Matthew Kelly
    Abstract. This paper is an introductory tutorial for numerical trajectory optimization with a focus on direct collocation methods.
  2. [2]
    Ch. 10 - Trajectory Optimization - Underactuated Robotics
    For trajectory optimization, we need a finite-dimensional parameterization in only one dimension (time), whereas in the mesh-based value iteration algorithms we ...
  3. [3]
    [PDF] Advances in trajectory optimization for space vehicle control
    Nov 2, 2021 · This section provides a broad overview of key algorithms for space vehicle trajectory optimization. The main focus is on methods that exploit ...<|control11|><|separator|>
  4. [4]
    Applied Optimal Control: Optimization, Estimation, and Control
    This best-selling text focuses on the analysis and design of complicated dynamics systems and is recommended by engineers, applied mathematicians, ...
  5. [5]
    Chapter 17. Optimal control
    Optimal control asks to compute a control function (either open loop or closed loop) that optimizes some performance metric regarding the control and the ...Optimal control problem · LQR control · Trajectory Optimization
  6. [6]
    [PDF] cvoc.pdf - Daniel Liberzon
    Aug 9, 2011 · We now have a refined formulation of the optimal control problem: given a control system (3.18) satisfying the assumptions of Section 3.3.1 ...
  7. [7]
    An Introduction to Trajectory Optimization: How to Do Your Own ...
    A problem with both terms is said to be in Bolza form. A problem with only the integral term is said to be in Lagrange form, and a problem with only a ...
  8. [8]
    Brachistochrone - MATHCURVE.COM
    Curve studied and named by Jean Bernoulli in 1718 and Euler in 1736. From ... The problem of the brachistochrone with given length is studied on this page.Missing: source | Show results with:source
  9. [9]
    Sir William Rowan Hamilton (1805-1865): Mathematical Papers
    This collection consists of the mathematical papers of Sir William Rowan Hamilton published during his lifetime, transcribed and edited by David R. Wilkins.
  10. [10]
    [PDF] The Early History of Hamilton-Jacobi Dynamics 1834–1837
    May 2, 2023 · The early history of Hamilton-Jacobi dynamics involves Hamilton's 1834-35 essays, Jacobi's 1836 letter and 1837 article, and his generalization ...
  11. [11]
    [PDF] A Brief Survey of the Calculus of Variations - arXiv
    In the calculus of variations, the Euler equation and the transversality conditions are among the so-called necessary conditions. They are derived by ...
  12. [12]
    [PDF] a method of reaching extreme - Smithsonian Institution
    The third additional calculations, Mr^, are carried out for the case of a rocket built up of Coston rockets in bundles (shown in section in fig. 22), the ...Missing: stage | Show results with:stage
  13. [13]
    [PDF] THE THEORY OF DYNAMIC PROGRAMMING - Richard Bellman
    stated above, the basic idea of the theory of dynamic programming is that of viewing an optimal policy as one deter- mining the decision required at each time ...Missing: work | Show results with:work
  14. [14]
    [PDF] DYNAMIC PROGRAMMING - Gwern
    RICHARD BELLMAN. PRINCETON UNIVERSITY PRESS. PRINCETON, NEW JERSEY. Page 3 ... PRINCIPLE OF OPTIMALITY. An optimal policy has the property that what- ever ...
  15. [15]
    [PDF] Pontryagin's Maximum Principle
    The Maximum Principle was formulated in 1956 by L.S. Pontryagin (Lev ... [This is the original (first) paper introducing the notion of controllability].
  16. [16]
    [PDF] NATIONAL u TECHNICAL INFORMATION SERVICE
    Optimal control of spacecraft trajectories requires obtaining an optimal (maximum or minimum) value for an appropriate scalar quantity. This scalar quantity ...
  17. [17]
    [PDF] Applied Optimal Control - Research - Free
    This text is intended for use at the senior-graduate level for uni- versity courses on the analysis and design of dynamic systems and for independent study by ...
  18. [18]
    [PDF] optimal control - theory - Free
    Kirk, Donald E., 1937–. Optimal control theory: an introduction / Donald E. Kirk. p. cm. Originally published: Englewood Cliffs, N.J.: Prentice-Hall, 1970 ( ...
  19. [19]
    A Legendre Pseudospectral Method for rapid optimization of launch ...
    The method uses a Legendre pseudospectral differentiation matrix to discretize nonlinear differential equations (such as the equations of motion) into ...Missing: 1990s Elishakoff
  20. [20]
    [PDF] Constrained Trajectory Optimization Using Pseudospectral Methods ...
    May 8, 2008 · As stated in the introduction, several numerical techniques exist to solve constrained two point boundary value problems cast as hypersonic ...Missing: Elishakoff | Show results with:Elishakoff
  21. [21]
    ACADO toolkit—An open‐source framework for automatic control ...
    May 17, 2011 · It provides a general framework for using a great variety of algorithms for direct optimal control, including model predictive control as well as state and ...
  22. [22]
    ACADO Toolkit
    It provides a general framework for using a great variety of algorithms for direct optimal control, including model predictive control, state and parameter ...ACADO from MATLAB · Features · Overview · Linux installationMissing: 2010s | Show results with:2010s
  23. [23]
    [PDF] C_-4501) TRAJECTQ_Y OPTIMIZATION _ASED ON DIFFERENTIAL ...
    The performance of the new algorithm is demonstrated on a 1-D rocket ascent problem ("Goddard Problem") in presence of a dynamic pressure constraint.
  24. [24]
    [PDF] Optimal Control Problems With Switching Points
    The Goddard. Problem is that of maximizing the final altitude for a vertically ascending, rocket-powered vehicle under the influence of an inverse square ...
  25. [25]
    Survey of Methods for Calculating Impulsive $\Delta V$-Minimizing ...
    Jan 5, 2020 · These solutions range from the Hohmann transfer to numerical optimization algorithms applied to finding a set of maneuvers that navigate between ...<|separator|>
  26. [26]
    A Class of Optimal Two-Impulse Rendezvous Using Multiple ...
    Aug 23, 2020 · Lambert's problem, the determination of the orbit that connects two specified terminal points in a specified time interval, affords multiple ...
  27. [27]
    [PDF] Mars Entry, Descent, and Landing Instrumentation 2 Trajectory ...
    On February 18th, 2021, the Mars 2020 entry system successfully delivered the Perseverance rover to the surface of Mars at Jezero Crater.
  28. [28]
    Postflight Assessment of Mars 2020 Entry, Descent, and Landing ...
    Feb 13, 2024 · The Program to Optimize Simulated Trajectories II, a trajectory simulation tool, was the prime entry, descent, and landing performance ...Missing: reentry | Show results with:reentry
  29. [29]
    [PDF] MINIMUM TIME AIRCRAFT TRAJECTORIES BETWEEN TWO ...
    The minimum time to climb problem is generally treated as a two point problem in velocity- altitude space. By definition the final range point for a time to ...
  30. [30]
    Robust Wind-Aware Path Optimization Onboard Small Fixed-wing ...
    Jan 19, 2023 · We propose a computationally lightweight path sampling algorithm to approximate the solution and enable direct integration with existing path following ...
  31. [31]
    [PDF] Apollo 11 Reloaded: Optimization-based Trajectory Reconstruction
    Jan 19, 2021 · By using modern methods based on numerical optimization we reconstruct critical phases of the original mission, and more specifically the ascent ...
  32. [32]
    [PDF] Optimization of Energy Consumption in KUKA KR 16 Articulated ...
    Sep 30, 2019 · Abstract: A study for optimal energy consumption in KUKA. KR 16 articulated robot for pick-and-place task was introduce in this paper.
  33. [33]
    [PDF] Trajectory Optimization for a 6 DOF Robotic Arm Based on ...
    Abstract: The design of the robotic arm's trajectory is based on inverse kinematics problem solving, with additional refinements of certain criteria.<|separator|>
  34. [34]
    [PDF] Fast Trajectory Optimization for Agile Quadrotor Maneuvers with a ...
    Tasks such as way- point navigation or obstacle avoidance often require a large number of parameters [25, 2] that must be hand-tuned by an expert operator and ...
  35. [35]
    Analysis and Insights from the DARPA Subterranean Challenge
    ... optimization methods to compute the robot's trajectory and map estimate [21]. ... Small quadrotor drones offer a potential solution, as they can navigate ...
  36. [36]
    [PDF] A closed-form solution for real-time ZMP gait generation and ...
    We describe an implementation on the. Boston Dynamics Atlas that demonstrates online ZMP re- planning and stabilization with sub-millisecond computation.
  37. [37]
    [PDF] 3D Walking Based on Online Optimization
    The proposed walking controller is tested on a simulated. Boston Dynamics' Atlas robot in DARPA's Virtual Robotics. Challenge setting. ... control of zero-moment ...
  38. [38]
    [PDF] Distributed Swarm Trajectory Optimization for Formation Flight in ...
    Apr 21, 2022 · This paper proposes an optimization method for collision-free formation flight, using a novel metric and a distributed framework to balance  ...
  39. [39]
    Optimal Control of a Swarming Multi-agent System through ...
    The leader tries by optimal interventions to steer the entire swarm towards objectives. For this purpose, we utilize an optimal control strategy. Possible ...
  40. [40]
    [PDF] Optimization-based walking generation for humanoid robot
    The purpose of this paper is to show that mathematical trajectory optimization or optimal control can be very helpful to generate walking motions for humanoid.<|control11|><|separator|>
  41. [41]
    [PDF] Unified Multi-Contact Fall Mitigation Planning for Humanoids ... - arXiv
    Oct 14, 2018 · The planner optimizes both the contact sequence and the robot state trajectories. A high-level tree search is conducted to iteratively grow a ...
  42. [42]
    Smooth time-optimal tool trajectory generation for CNC ...
    Technical paper. Smooth time-optimal tool trajectory generation for CNC manufacturing systems.
  43. [43]
    Versatile modeling and optimization of fed batch processes for the ...
    Dec 11, 2006 · Fermentation temperature was controlled at 25°C, pH was controlled at 5.0 with addition of 25% ammonium hydroxide and the dissolved oxygen ...
  44. [44]
    Optimal control of a batch bioreactor: a study on the use of an ...
    This paper presents a numerical study on the performance degradation of a batch bioreactor optimal control strategy when the model used to compute such ...Optimal Control Of A Batch... · Introduction · Penicillin Batch...<|separator|>
  45. [45]
    Autonomous Vehicle Fuel Economy Optimization with Deep ... - MDPI
    Thus, we propose a deep reinforcement learning-based algorithm that trains autonomous vehicles to optimize fuel while following a given trajectory.Autonomous Vehicle Fuel... · 2.1. Fuel Economy Simulation... · 3.2. Deep Neural Network...
  46. [46]
    Optimization for total energy consumption of drone inspection based ...
    Dec 1, 2024 · This paper focuses on the TIEC optimization of UAV automatic inspection of offshore wind farms based on the assistance of mobile edge computing (MEC).
  47. [47]
    Optimal control of rapid thermal annealing in a semiconductor process
    This study focuses on the optimal control of rapid thermal annealing (RTA) used in the formation of ultrashallow junctions needed in next-generation ...
  48. [48]
    [PDF] Optimization-Based Control
    Our focus is on the use of optimization-based methods for control, including optimal control theory, receding horizon control and Kalman filtering. Each chapter ...<|control11|><|separator|>
  49. [49]
  50. [50]
  51. [51]
    [PDF] Optimal Control under State Constraints - IMJ-PRG
    Abstract. Optimal control under state constraints has brought new mathematical chal- lenges that have led to new techniques and new theories.
  52. [52]
  53. [53]
  54. [54]
    [PDF] GLOBAL, MULTI-OBJECTIVE TRAJECTORY OPTIMIZATION WITH ...
    This paper addresses multi-objective trajectory optimization, using a genetic algorithm to generate Pareto solutions, and addresses clustering by using ...
  55. [55]
    [PDF] An Introduction to Mathematical Optimal Control Theory Spring ...
    Let α∗(·) be an optimal control, corresponding to the optimal trajectory x∗(·). Our plan is to compute “variations” of the optimal control and to use an adjoint.
  56. [56]
    Survey of Numerical Methods for Trajectory Optimization - AIAA ARC
    Survey of Numerical Methods for Trajectory Optimization. John T. Betts. John T. Betts. Boeing Information and Support Services, Seattle, Washington 98124- ...
  57. [57]
    The mathematical theory of optimal processes - Internet Archive
    Jun 20, 2019 · The mathematical theory of optimal processes. by: Pontri͡agin, L. S. (Lev Semenovich), 1908-. Publication date: 1962. Topics: Mathematical ...
  58. [58]
    [PDF] Pontryagin's maximum principle and indirect methods
    May 3, 2023 · The boundary condition h(x∗(T),p∗(T)) = 0 is determined by the terminal set constraint x∗(T) ∈ XT and the transversality condition −p∗(T) − η ∇ ...
  59. [59]
    [PDF] Transcription Methods for Trajectory Optimization - arXiv
    Jul 2, 2017 · Multiple shooting works by breaking up a trajectory into some number of segments, and using single shooting to solve for each segment. As the ...
  60. [60]
    A review of pseudospectral optimal control: From theory to flight
    In this paper, we review key theoretical results in pseudospectral optimal control that have proven to be critical for a successful flight.
  61. [61]
    [PDF] High order variational integrators in the optimal control of ... - arXiv
    Feb 1, 2015 · In this article, we develop two high order variational integrators which distinguish themselves in the dimension of the underling space of ...
  62. [62]
    [PDF] Model-Based Diffusion for Trajectory Optimization - NIPS papers
    Our approach uses diffusion models directly as solvers, rather than simply distilling solutions from demonstrations. Langevin-based Markov Chain Monte Carlo for ...Missing: 2020s | Show results with:2020s
  63. [63]
  64. [64]
  65. [65]
    An hp‐adaptive pseudospectral method for solving optimal control ...
    Aug 25, 2010 · An hp-adaptive pseudospectral method is presented for numerically solving optimal control problems. The method presented in this paper iteratively determines ...
  66. [66]
    [PDF] An hp-adaptive pseudospectral method for solving optimal control ...
    Oct 6, 2009 · SUMMARY. An hp-adaptive pseudospectral method is presented for numerically solving optimal control problems. The method presented in this ...
  67. [67]
    [PDF] Direct Trajectory Optimization and Costate Estimation of Finite ...
    It is noted that some researchers prefer the term orthogonal collocation,17–19 but the terms pseudospectral and orthogonal collocation have the same meaning.
  68. [68]
    [PDF] Direct Trajectory Optimization by a Chebyshev Pseudospectral Method
    A Chebyshev pseudospectral method is presented in this paper for directly solving a generic optimal con- trol problem with state and control constraints.<|separator|>
  69. [69]
    Direct multiple shooting for computationally efficient train trajectory ...
    An intrinsic disadvantage of the pseudospectral method against direct collocation or multiple shooting is the reduced sparsity in the discretized problem, ...
  70. [70]
    Pseudospectral Optimal Control Theory Makes Debut Flight, Saves ...
    PDF | This paper describes the use of a feedforward fuel-optimal trajectory to command the International Space Station (ISS) and perform a large angle.
  71. [71]
    Combining trajectory optimization, supervised machine learning ...
    Jul 8, 2019 · This is where the “curse of dimensionality” rears its ugly head. Placing 10 points per dimension leads to 10 n optimizations to compute, which ...
  72. [72]
    Combining Trajectory Optimization, Supervised Machine Learning ...
    Nov 6, 2017 · Combining Trajectory Optimization, Supervised Machine Learning, and Model Structure for Mitigating the Curse of Dimensionality in the Control of ...
  73. [73]
    [PDF] Warm Start of Mixed-Integer Programs for Model Predictive Control ...
    The main idea behind it is straightforward: if we are able to solve trajectory optimization problems quickly enough, we can replan the future motion of the ...
  74. [74]
    [PDF] Fast Model Predictive Control Using Online Optimization
    In this case, warm starting from the previously computed trajectory reliably gives a very good advantage in terms of the number of Newton steps required. In ...<|separator|>
  75. [75]
  76. [76]
    [PDF] Chance-Constrained Sequential Convex Programming for Robust ...
    Abstract— Planning safe trajectories for nonlinear dynamical systems subject to model uncertainty and disturbances is challenging.
  77. [77]
    [PDF] Robust Trajectory Optimization Applying Chance Constraints and ...
    Here, the presence of uncertainties normally requires a trade-off between the optimality of the cost function and the robustness, i.e., the reduction of the ...
  78. [78]
    Chance-Constraint Robust Trajectory Optimization with a Hybrid ...
    Sep 21, 2025 · This paper proposes a systematic, robust trajectory design method, where quantitative information concerning uncertainty on the system dynamics ...
  79. [79]
    [PDF] Large Scale Constrained Trajectory Optimization Using Indirect ...
    Survey of numerical methods for trajectory optimization. Journal of Guidance, Control and Dynamics, 21(2):193–207, 1998. [9] F Fahroo, DB Doman, and AD Ngo ...
  80. [80]
    Distributed safe trajectory optimization for large-scale spacecraft ...
    This paper presents a distributed trajectory optimization algorithm for the reconfiguration of large-scale spacecraft formation.
  81. [81]
    Optimizing Guidance and Control Networks through Neural ODEs
    Apr 25, 2024 · We demonstrate that for the orbital transfer, the final error to the target can be reduced by 99% on a single trajectory and by 70% on a batch ...
  82. [82]
    Aligning Optimization Trajectories with Diffusion Models for ... - arXiv
    May 29, 2023 · The paper introduces Diffusion Optimization Models (DOM) and Trajectory Alignment (TA), aligning diffusion model sampling with physics-based  ...
  83. [83]
    Parallelized Particle Swarm Optimization on FPGA for Realtime ...
    Oct 13, 2023 · This paper proposes a parallelized PSO technique using FPGA to accelerate real-time ballistic target tracking, reducing computation time by up ...
  84. [84]
    Online Trajectory Optimization Method for Large Attitude Flip Vertical ...
    For this problem size, GPOPS-II is invoked to solve the nonconvex fuel optimal trajectory optimization problem P 1 , the augmented nonconvex fuel optimal ...
  85. [85]
    Research and Simulation Implementation of Self-Driving Vehicle ...
    Feb 21, 2025 · It is found that the trajectory optimization scheme not only improves the traffic condition of the intersection, but also reduces the average CO ...Missing: economy | Show results with:economy
  86. [86]