What question did this study set out to answer?

The research aims to enhance the interpretability of reinforcement learning policies through improved trajectory optimization.

April 1, 2026Open Access

Surrogate Fitness Metrics for Interpretable Reinforcement Learning

Key Points

The research aims to enhance the interpretability of reinforcement learning policies through improved trajectory optimization.
Utilized an evolutionary optimization framework to perturb initial states for RL policy demonstrations.
Developed a surrogate fitness function combining local diversity, behavioral certainty, and global population diversity.
Assessed demonstration quality using evaluation metrics like reward-based optimality gap and fidelity interquartile means.
Examined hyperparameter sensitivity to understand trajectory optimization dynamics.
Optimizing trajectory selection via surrogate fitness metrics significantly improved interpretability of RL policies.
Demonstration fidelities in gridworld domains were enhanced compared to random and ablated baselines.
The framework provided valuable insights for early-stage RL policies, while fidelity-based optimization was more effective for matured policies.

Abstract

We employ an evolutionary optimization framework that perturbs initial states to generate informative and diverse reinforcement learning (RL) policy demonstrations. A surrogate fitness function guides the optimization by combining local diversity, behavioral certainty, and global population diversity. To assess demonstration quality, we apply a set of evaluation metrics, including the reward-based optimality gap, fidelity interquartile means (IQMs), fitness composition analysis, and trajectory visualizations. Hyperparameter sensitivity is also examined to better understand the dynamics of trajectory optimization. Our findings demonstrate that optimizing trajectory selection via surrogate fitness metrics significantly improves the interpretability of RL policies in both discrete and continuous environments. In gridworld domains, evaluations reveal significantly enhanced demonstration fidelities compared to random and ablated baselines. In continuous control, the proposed framework provides valuable insights, particularly for early-stage policies, whereas fidelity-based optimization is more effective for mature policies. By refining and systematically analyzing surrogate fitness functions, this study advances the interpretability of RL models. The proposed improvements provide deeper insights into RL decision-making, benefiting applications in safety-critical and explainability-focused domains.

Bookmark

View Full Paper