Enhancing Diffusion Policy with Classifier-Free Guidance for Temporal Robotic Tasks

2025-10-10

AI Summary

Diffusion policies (DP) and Action Chunking with Transformers (ACT) suffer from local optima, action repetition, and unreliable termination in humanoid robot time-series tasks due to the absence of explicit temporal modeling. To address these issues, we propose a temporally aware diffusion policy framework integrated with classifier-free guidance (CFG). Our method conditions the diffusion process on discrete timesteps, enabling dynamic modulation of action prediction across the trajectory; a tunable CFG scale factor jointly optimizes conditional and unconditional model outputs, thereby enhancing temporal coherence and phase-awareness without compromising action accuracy. Experiments on a real-world humanoid robot platform demonstrate that our approach significantly improves task success rates, effectively suppresses redundant actions, achieves precise periodic termination, and exhibits strong robustness under varying environmental conditions.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Computer Vision: Diffusion Models for Vision
• Intelligent Robots: Manipulation

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Temporal sequential tasks challenge humanoid robots, as existing Diffusion Policy (DP) and Action Chunking with Transformers (ACT) methods often lack temporal context, resulting in local optima traps and excessive repetitive actions. To address these issues, this paper introduces a Classifier-Free Guidance-Based Diffusion Policy (CFG-DP), a novel framework to enhance DP by integrating Classifier-Free Guidance (CFG) with conditional and unconditional models. Specifically, CFG leverages timestep inputs to track task progression and ensure precise cycle termination. It dynamically adjusts action predictions based on task phase, using a guidance factor tuned to balance temporal coherence and action accuracy. Real-world experiments on a humanoid robot demonstrate high success rates and minimal repetitive actions. Furthermore, we assessed the model's ability to terminate actions and examined how different components and parameter adjustments affect its performance. This framework significantly enhances deterministic control and execution reliability for sequential robotic tasks.

Problem

• Addressing temporal context gaps in robot sequential tasks
• Reducing local optima traps and repetitive robot actions
• Enhancing deterministic control and execution reliability

Innovation

• Classifier-Free Guidance enhances Diffusion Policy framework
• Timestep inputs track progression and ensure cycle termination
• Guidance factor balances temporal coherence with action accuracy