Online, Model-Free Motion Planning in Dynamic Environments: An Intermittent, Finite Horizon Approach with Continuous-Time Q-Learning.
George P. KontoudisZirui XuKyriakos G. VamvoudakisPublished in: ACC (2020)
Keyphrases
- dynamic environments
- model free
- motion planning
- finite horizon
- path planning
- mobile robot
- reinforcement learning
- trajectory planning
- optimal policy
- potential field
- infinite horizon
- reinforcement learning algorithms
- collision free
- markov decision processes
- policy iteration
- optimal control
- function approximation
- state space
- robotic tasks
- belief space
- degrees of freedom
- markov decision process
- collision avoidance
- temporal difference
- multi robot
- multistage
- obstacle avoidance
- real time
- robotic arm
- humanoid robot
- average reward
- autonomous robots
- rl algorithms
- markov chain
- path finding
- non stationary
- stochastic processes
- average cost
- partially observable
- finite state
- robotic systems