End-To-End Robotic Reinforcement Learning without Reward Engineering.
Avi SinghLarry YangChelsea FinnSergey LevinePublished in: Robotics: Science and Systems (2019)
Keyphrases
- end to end
- reinforcement learning
- real robot
- eligibility traces
- wireless ad hoc networks
- state space
- reinforcement learning algorithms
- admission control
- ad hoc networks
- mobile robot
- multipath
- markov decision processes
- congestion control
- learning algorithm
- model free
- reward function
- real time
- learning agent
- content delivery
- high bandwidth
- transport layer
- optimal policy
- average reward
- policy gradient