Learning From Failures: Efficient Reinforcement Learning Control with Episodic Memory

2026-03-07

AI Summary

In contact-rich and exploration-challenging robotic tasks, reinforcement learning often suffers from numerous short-sighted failure trajectories caused by frequent early collisions or falls, which hinder policy convergence and long-term exploration. To address this issue, this work proposes Failure Episodic Memory Alert (FEMA), a novel approach that, for the first time, integrates an episodic memory-based mechanism for storing and retrieving failure experiences into model-free reinforcement learning. FEMA explicitly prevents the policy from revisiting unstable states, thereby guiding learning toward high-reward trajectories. The method seamlessly integrates with algorithms such as PPO and supports parallelized training. Empirical evaluations on MuJoCo benchmarks demonstrate an average improvement of 33.11% in sample efficiency, and real-world experiments on a bipedal robot further validate its effectiveness.

Technology Category

• Intelligent Robots: Learning & Optimization for ROB
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Multiagent Systems: Multiagent Learning

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Reinforcement learning has achieved remarkable success in robot learning. However, under challenging exploration and contact-rich dynamics, early-stage training is frequently dominated by premature terminations such as collisions and falls. As a result, learning is overwhelmed by short-horizon, low-return trajectories, which hinder convergence and limit long-horizon exploration. To alleviate this issue, we propose a technique called Failure Episodic Memory Alert (FEMA). FEMA explicitly stores short-horizon failure experiences through an episodic memory module. During interactions, it retrieves similar failure experiences and prevents the robot from recurrently relapsing into unstable states, guiding the policy toward long-horizon trajectories with greater long-term value. FEMA can be combined easily with model-free reinforcement learning algorithms, and yields a substantial sample-efficiency improvement of 33.11% on MuJoCo tasks across several classical RL algorithms. Furthermore, integrating FEMA into a parallelized PPO training pipeline demonstrates its effectiveness on a real-world bipedal robot task.

Problem

• reinforcement learning
• premature termination
• short-horizon trajectories
• exploration
• contact-rich dynamics

Innovation

• episodic memory
• reinforcement learning
• failure recovery
• sample efficiency
• robot control