SAT-MARL: Specification Aware Training in Multi-Agent Reinforcement Learning

12/14/2020

∙

A characteristic of reinforcement learning is the ability to develop unforeseen strategies when solving problems. While such strategies sometimes yield superior performance, they may also result in undesired or even dangerous behavior. In industrial scenarios, a system's behavior also needs to be predictable and lie within defined ranges. To enable the agents to learn (how) to align with a given specification, this paper proposes to explicitly transfer functional and non-functional requirements into shaped rewards. Experiments are carried out on the smart factory, a multi-agent environment modeling an industrial lot-size-one production facility, with up to eight agents and different multi-agent reinforcement learning algorithms. Results indicate that compliance with functional and non-functional constraints can be achieved by the proposed approach.

READ FULL TEXT

SAT-MARL: Specification Aware Training in Multi-Agent Reinforcement Learning

K-nearest Multi-agent Deep Reinforcement Learning for Collaborative Tasks with a Variable Number of Agents

Multi-Agent Deep Reinforcement Learning with Human Strategies

Policy-Based Reinforcement Learning for Assortative Matching in Human Behavior Modeling

Scalable Evaluation of Multi-Agent Reinforcement Learning with Melting Pot

Parameter Sharing Reinforcement Learning Architecture for Multi Agent Driving Behaviors

Hierarchical Needs-driven Agent Learning Systems: From Deep Reinforcement Learning To Diverse Strategies

A General Framework for Interacting Bayes-Optimally with Self-Interested Agents using Arbitrary Parametric Model and Model Prior

SAT-MARL: Specification Aware Training in Multi-Agent Reinforcement Learning

Related Research

K-nearest Multi-agent Deep Reinforcement Learning for Collaborative Tasks with a Variable Number of Agents

Multi-Agent Deep Reinforcement Learning with Human Strategies

Policy-Based Reinforcement Learning for Assortative Matching in Human Behavior Modeling

Scalable Evaluation of Multi-Agent Reinforcement Learning with Melting Pot

Parameter Sharing Reinforcement Learning Architecture for Multi Agent Driving Behaviors

Hierarchical Needs-driven Agent Learning Systems: From Deep Reinforcement Learning To Diverse Strategies

A General Framework for Interacting Bayes-Optimally with Self-Interested Agents using Arbitrary Parametric Model and Model Prior