Incorporation of Military Doctrines and Objectives into an AI Agent Via Natural Language and Reward in Reinforcement Learning.
Michael MöbiusDaniel KallfassMatthias FlockThomas DollDietmar KundePublished in: WSC (2023)
Keyphrases
- reinforcement learning
- natural language
- learning agent
- reward function
- multi agent
- machine learning
- action selection
- agent receives
- knowledge representation
- state action
- partially observable
- state space
- multi agent systems
- artificial intelligence
- multi agent environments
- exploration strategy
- reinforcement learning algorithms
- state abstraction
- reward shaping
- agent learns
- expert systems
- single agent
- markov decision processes
- autonomous agents
- function approximation
- multiagent systems
- autonomous learning
- learning capabilities
- reward signal
- model free
- eligibility traces
- case based reasoning
- markov decision process
- software agents
- action space
- average reward
- learning agents
- inverse reinforcement learning
- reinforcement learning agents
- multiple agents
- learning algorithm
- intelligent agents
- agent model
- dynamic environments
- optimal policy
- total reward
- expected reward
- learning process
- intelligent systems
- temporal difference
- decision making
- optimal control
- control policy
- intelligent behavior
- robocup soccer
- agent architecture
- agent technology
- dynamic programming
- partially observable environments
- question answering
- natural language processing
- mobile agents
- policy gradient
- learning tasks
- solving problems