Robot Fine-Tuning Made Easy: Pre-Training Rewards and Policies for Autonomous Real-World Reinforcement Learning.
Jingyun YangMax Sobol MarkBrandon VuArchit SharmaJeannette BohgChelsea FinnPublished in: ICRA (2024)
Keyphrases
- fine tuning
- reinforcement learning
- real robot
- real world
- autonomous learning
- optimal policy
- reward function
- fine tuned
- robotic systems
- markov decision processes
- semi autonomous
- robot behavior
- supervised learning
- mobile robot
- total reward
- function approximation
- policy search
- perceptual aliasing
- viable alternative
- robot control
- fine tune
- control policy
- control policies
- data sets
- markov decision problems
- reinforcement learning algorithms
- service robots
- autonomous navigation
- fitted q iteration
- markov decision process
- learning capabilities
- training set
- reinforcement learning agents
- dynamic programming
- state space
- partially observable markov decision processes
- real environment
- vision system
- multi robot
- machine learning
- hierarchical reinforcement learning
- human robot interaction
- decentralized control
- learning algorithm
- multi agent
- path planning
- search and rescue
- humanoid robot
- model free
- robot navigation
- temporal difference
- motor skills
- infinite horizon
- robot arm
- action selection
- reward shaping
- autonomous vehicles