Advising reinforcement learning toward scaling agents in continuous control environments with sparse rewards.
Hailin RenPinhas Ben-TzviPublished in: Eng. Appl. Artif. Intell. (2020)
Keyphrases
- reinforcement learning
- multi agent
- action selection
- control problems
- multi agent environments
- markov decision processes
- action space
- control policy
- learning agents
- dynamic environments
- robot control
- multi agent systems
- autonomous agents
- learning agent
- state space
- function approximation
- optimal control
- single agent
- reinforcement learning agents
- control strategies
- agent behavior
- intelligent agents
- decentralized control
- multiple agents
- highly dynamic
- partial observability
- reward signal
- continuous state and action spaces
- cooperative
- learning algorithm
- agent receives
- complex domains
- robotic systems
- multiagent systems
- robocup soccer
- control system
- multi agent reinforcement learning
- multiagent learning
- autonomous systems
- complex environments
- machine learning
- partially observable
- agent architecture
- reward function
- optimal policy
- distributed control
- software agents
- agent learns
- reward shaping
- model free
- open systems
- learning capabilities
- autonomous robots
- dynamic programming
- function approximators
- high dimensional
- adjustable autonomy
- agent teams
- control strategy