RAPAC-DP: Response-Aligned Pending-Action Compensation for Diffusion Policies under Delayed Execution

2026-08-16

AI Summary

This study addresses the performance degradation in imitation learning caused by communication and computational latencies during cloud-based inference. We propose a Response-Aligned Pending Action Compensation framework that encodes scheduled action sequences as conditional inputs to achieve delay compensation via parameter-efficient pathways without explicit dynamics modeling. This approach seamlessly restores baseline policy performance without requiring additional data. Experimental results demonstrate that the framework retains 81.4% of performance under maximum latency in Kinetix and achieves an average success rate of 0.633 across three RoboMimic tasks. These findings confirm the method’s effectiveness in resolving temporal misalignment challenges inherent to cloud-edge collaborative inference systems.

Abstract

Cloud-side inference gives imitation-learning policies access to greater computational resources, but communication and computation delays can degrade control performance. To compensate for these delays, we propose RAPAC-DP, a response-aligned pending-action compensation framework designed for both diffusion- and flow-based action generators. RAPAC-DP encodes the actions already scheduled for execution before the cloud response arrives into a pending-action sequence that serves as the conditioning input to a parameter-efficient compensation pathway. When delay effects are negligible, bypassing this pathway exactly recovers the frozen base policy. For training, RAPAC-DP constructs delay-conditioned samples from delay-free demonstrations, requiring neither explicit system dynamics nor additional delayed demonstrations. At the largest fixed delay tested on Kinetix, RAPAC-DP retained 81.4% of its overall delay-free performance. At the largest fixed delay tested on each RoboMimic task, it achieved a mean success rate of 0.633 across the three tasks. These results demonstrate the effectiveness of pending-action compensation for cloud-deployed imitation-learning policies.

Problem

• Delayed Execution
• Cloud-side Inference
• Imitation Learning
• Diffusion Policies

Innovation

• Pending-Action Compensation
• Diffusion Policy
• Delayed Execution
• Parameter-Efficient Adaptation
• Cloud-Side Inference