Enhancing RL Generalizability in Robotics through SHAP Analysis of Algorithms and Hyperparameters

2026-05-04

AI Summary

This work addresses the limited generalization of reinforcement learning in robotic applications, often hindered by the sensitivity of algorithms and hyperparameters to environmental variations, which complicates real-world deployment. To tackle this challenge, the paper introduces the first explainable framework based on SHAP (Shapley Additive exPlanations) that systematically quantifies the contribution of each configuration—such as algorithmic choices and hyperparameters—to the generalization gap. It establishes a theoretical link between these contributions and generalization performance and leverages this insight to design a SHAP-guided configuration selection strategy. Empirical evaluations across multi-task and cross-environment settings demonstrate that the proposed approach significantly enhances model generalization, uncovers stable patterns in how configurations affect performance, and offers practitioners actionable, data-driven guidance for selecting algorithms and hyperparameters.

Abstract

Despite significant advances in Reinforcement Learning (RL), model performance remains highly sensitive to algorithm and hyperparameter configurations, while generalization gaps across environments complicate real-world deployment. Although prior work has studied RL generalization, the relative contribution of specific configurations to the generalization gap has not been quantitatively decomposed and systematically leveraged for configuration selection. To address this limitation, we propose an explainable framework that evaluates RL performance across robotic environments using SHapley Additive exPlanations (SHAP) to quantify configuration impacts. We establish a theoretical foundation connecting Shapley values to generalizability, empirically analyze configuration impact patterns, and introduce SHAP-guided configuration selection to enhance generalization. Our results reveal distinct patterns across algorithms and hyperparameters, with consistent configuration impacts across diverse tasks and environments. By applying these insights to configuration selection, we achieve improved RL generalizability and provide actionable guidance for practitioners.

Problem

• Reinforcement Learning
• Generalization
• Hyperparameters
• Robotics
• SHAP

Innovation

• SHAP analysis
• Reinforcement Learning generalization
• hyperparameter selection
• explainable AI
• robotic environments