Robot learning system saves memory while avoiding forgetting
RIFAR: Reliability and Forgetting-Aware Replay for Continual Robot Learning
Artificial Intelligence
Summary
Robots need to learn new tasks without forgetting how to do previous ones, but keeping all their past experiences takes a lot of memory. The authors developed RIFAR, a method that stores small parts of past experiences and checks if predicted robot actions match what actually happens. It keeps the most reliable experiences and adapts its memory to avoid losing old skills while learning new ones. This approach uses much less memory yet keeps robot learning effective over time.
What this means in practice
- •For robot developers: Integrate reliable compact replay methods to maintain multiple robotic skills with minimal memory use in continuous learning.
- •For industrial automation teams: Deploy robots that can learn new tasks on factory floors without losing previous capabilities while using limited data storage.
Authors
Zirong Song, Zheng Lu, Haoran Liao, Wanqi Zhong, Yunhe Ni, Lijie Wang, Xiuying Chen
Abstract
Genuine embodied agency requires robots to turn continuous real-world experience into lasting, transferable skills. This demands continual learning that integrates new capabilities without eroding prior knowledge as tasks and environments evolve. Experience replay mitigates forgetting, but storing complete demonstrations becomes costly as tasks accumulate. World-action models offer a generative alternative, reconstructing past experience through joint predictions of actions and future observations. However, visually coherent rollouts may contain actions that cannot realize the predicted transitions, while new-task adaptation can disrupt previously learned behavior. RIFAR therefore combines reliability screening with drift-aware replay selection. It reconstructs trajectories from compact demonstration prefixes and uses a frozen inverse-dynamics model to assess action-visual consistency. Training first combines current demonstrations with the highest-quality screened trajectories. RIFAR then compares action predictions before and after this adaptation on identical historical inputs, reselecting trajectories with larger normalized drift from the same screened pool for continued training. Across three LIBERO suites and real-world experiments, RIFAR surpasses the previous state of the art in WAM-based generative replay. On LIBERO-Goal, it achieves 90.97 AUC while retaining only 320 historical time steps per task, approximately 4.9% of the steps retained using 50-demonstration replay.