Overcoming Obstacles With a Reconfigurable Robot Using Deep Reinforcement Learning Based on a Mechanical Work-Energy Reward Function.
Or SimhonZohar KarniSigal BermanDavid ZarroukPublished in: IEEE Access (2023)
Keyphrases
- reward function
- reinforcement learning
- reinforcement learning algorithms
- robot moves
- markov decision processes
- mobile robot
- state space
- optimal policy
- real robot
- policy search
- partially observable
- collision free
- initially unknown
- learning agent
- robot control
- multiple agents
- reward signal
- autonomous robots
- markov decision process
- inverse reinforcement learning
- agent learns
- multi robot
- transition model
- path planning
- function approximation
- robot navigation
- model free
- state variables
- robotic systems
- hierarchical reinforcement learning
- temporal difference
- action selection
- humanoid robot
- multi agent
- dynamic environments
- learning algorithm
- image segmentation
- bayesian networks
- semi supervised
- dynamic systems
- control policies
- state action
- function approximators
- machine learning