Safety Correction from Baseline: Towards the Risk-aware Policy in Robotics via Dual-agent Reinforcement Learning.
Linrui ZhangZichen YanLi ShenShoujie LiXueqian WangDacheng TaoPublished in: IROS (2022)
Keyphrases
- reinforcement learning
- action selection
- reward function
- markov decision process
- agent learns
- optimal policy
- state action
- agent receives
- partially observable
- multi agent
- state space
- policy search
- action space
- decision making
- learning agent
- temporal difference
- inverse reinforcement learning
- reinforcement learning algorithms
- markov decision processes
- function approximation
- exploration strategy
- artificial intelligence
- partially observable environments
- control policy
- markov decision problems
- agent architecture
- state and action spaces
- selective perception
- partially observable markov decision process
- learning algorithm
- multi agent environments
- computer vision
- multiagent systems
- policy gradient
- multiple agents
- real robot
- robot control
- reward shaping
- multi agent systems
- autonomous agents
- function approximators
- control policies
- partially observable domains
- robotic systems
- reward signal
- agent model
- learning capabilities
- learning process
- state abstraction
- dynamic programming
- mobile agents
- autonomous learning
- intelligent agents
- continuous state
- software agents
- learning agents
- average reward
- risk management
- partially observable markov decision processes
- long run
- expected reward
- machine learning
- traffic signal
- reinforcement learning problems
- mobile robot
- continuous state spaces
- rl algorithms
- reinforcement learning methods
- decision problems
- risk assessment
- decision theoretic
- infinite horizon