Robot control
Kinodynamic motion planning
Definition
Kinodynamic motion planning finds controls and a time-evolving robot trajectory that avoid obstacles while satisfying both kinematic and dynamic constraints. The planned state normally includes quantities such as velocity that a purely geometric path can omit.
Also known as: Kinodynamic planning, Kinodynamic trajectory planning
Updated
Planning motion that a robot can execute
A geometric planner can return a collision-free sequence of configurations without deciding how quickly the robot moves between them. Kinodynamic planning searches a state space that can include configuration, velocity, and other dynamic variables. Its edges are generated by admissible controls and the equations of motion.
The goal can specify both a final pose and velocity. Along the way, the trajectory must respect obstacles plus limits on quantities such as acceleration, steering rate, torque, or momentum. LaValle and Kuffner formalised this as planning under differential and obstacle constraints in a higher-dimensional state space.
Uses in dynamic robot motion
The method matters when inertia or nonholonomic motion cannot be ignored. Examples include a mobile robot that cannot move sideways, a manipulator intercepting a moving object, and a legged robot performing a fast manoeuvre. A rapidly-exploring random tree can be adapted by sampling controls and propagating the dynamic model instead of connecting arbitrary configurations with straight lines. LaValle's RRT publication record identifies the 2001 randomised kinodynamic planner and later work on steering under differential constraints.
Kinodynamic planning is narrower than motion planning, which also includes geometric methods. It is related to trajectory optimisation, but a sampling-based kinodynamic planner explores reachable states while a trajectory optimiser usually improves a parameterised candidate. Hybrid systems can use one to initialise the other.
Dynamics increase the search burden
Adding velocity and other state variables increases dimension. Many nonlinear systems lack an exact steering function that can connect two sampled states, so planners must choose control samples and integration durations. Narrow passages in state space and long dynamic horizons can require many rollouts.
Feasibility also depends on model fidelity. Unmodelled contact, actuator delay, friction, payload, or state-estimation error can make a planned trajectory unsafe or unreachable on hardware. Collision checking between integration steps, closed-loop tracking, uncertainty margins, and replanning remain necessary. A trajectory that satisfies a simplified model is not proof that the physical robot can execute it.
Sources
Related terms
Motion planning
Motion planning finds a robot movement from an initial state to a goal while satisfying constraints such as collision avoidance. A planner may produce a geometric path, a timed trajectory, or a sequence of controls.
Rapidly-exploring random tree
A rapidly-exploring random tree is a sampling-based structure that grows through a configuration or state space toward sampled targets. Motion planners use it to search for feasible routes through spaces with obstacles and movement constraints.
Trajectory optimization
Trajectory optimization finds a time-varying motion, and often control inputs, that minimizes an objective while satisfying specified constraints. Robot applications can include geometric, kinematic, and dynamic constraints.
Configuration space
Configuration space is the set of all possible configurations of a robot or mechanical system. Each point specifies the entire modeled arrangement, and the space has as many local dimensions as the system has degrees of freedom.