Timezone: »

Value Function Decomposition for Iterative Design of Reinforcement Learning Agents
James MacGlashan · Evan Archer · Alisa Devlic · Takuma Seno · Craig Sherstan · Peter Wurman · Peter Stone

Tue Nov 29 02:00 PM -- 04:00 PM (PST) @ Hall J #511

Designing reinforcement learning (RL) agents is typically a difficult process that requires numerous design iterations. Learning can fail for a multitude of reasons and standard RL methods provide too few tools to provide insight into the exact cause. In this paper, we show how to integrate \textit{value decomposition} into a broad class of actor-critic algorithms and use it to assist in the iterative agent-design process. Value decomposition separates a reward function into distinct components and learns value estimates for each. These value estimates provide insight into an agent's learning and decision-making process and enable new training methods to mitigate common problems. As a demonstration, we introduce SAC-D, a variant of soft actor-critic (SAC) adapted for value decomposition. SAC-D maintains similar performance to SAC, while learning a larger set of value predictions. We also introduce decomposition-based tools that exploit this information, including a new reward \textit{influence} metric, which measures each reward component's effect on agent decision-making. Using these tools, we provide several demonstrations of decomposition's use in identifying and addressing problems in the design of both environments and agents. Value decomposition is broadly applicable and easy to incorporate into existing algorithms and workflows, making it a powerful tool in an RL practitioner's toolbox.

Author Information

James MacGlashan (Sony AI)
Evan Archer (Sony AI)
Alisa Devlic (Sony AI)
Takuma Seno (Sony AI)
Craig Sherstan (Sony AI)
Peter Wurman (North Carolina State University)
Peter Stone (The University of Texas at Austin, Sony AI)

More from the Same Authors