Sparse Diffusion Policy: A Sparse, Reusable, and Flexible Policy for Robot Learning

2024/07/01 by Yixiao Wang, Yifei Zhang, Wang, Yixiao +19 · 19 citations
Computer Science · #FOS: Computer and information sciences #Machine Learning (cs.LG) #Reinforcement Learning in Robotics #Robotics (cs.RO)

paper · pdf · doi:10.48550/arxiv.2407.01531

openalex publication_date 2024/07/01 · openalex created_date 2024/07/06 · openalex updated_date 2026/07/28

Abstract

The increasing complexity of tasks in robotics demands efficient strategies for multitask and continual learning. Traditional models typically rely on a universal policy for all tasks, facing challenges such as high computational costs and catastrophic forgetting when learning new tasks. To address these issues, we introduce a sparse, reusable, and flexible policy, Sparse Diffusion Policy (SDP). By adopting Mixture of Experts (MoE) within a transformer-based diffusion policy, SDP selectively activates experts and skills, enabling efficient and task-specific learning without retraining the entire model. SDP not only reduces the burden of active parameters but also facilitates the seamless integration and reuse of experts across various tasks. Extensive experiments on diverse tasks in both simulations and real world show that SDP 1) excels in multitask scenarios with negligible increases in active parameters, 2) prevents forgetting in continual learning of new tasks, and 3) enables efficient task transfer, offering a promising solution for advanced robotic applications. Demos and codes can be found in https://forrest-110.github.io/sparse_diffusion_policy/.

Cited by

• Embodied Robot Manipulation in the Era of Foundation Models: Planning and Learning Perspectives
• Flexible Multitask Learning with Factorized Diffusion Policy
• Learning Semantic Atomic Skills for Multi-Task Robotic Manipulation
• Kinematics-Aware Diffusion Policy with Consistent 3D Observation and Action Space for Whole-Arm Robotic Manipulation
• Embodied Co-Design for Rapidly Evolving Agents: Taxonomy, Frontiers, and Challenges
• L1 Sample Flow for Efficient Visuomotor Learning
• Embodiment Transfer Learning for Vision-Language-Action Models
• Language-Conditioned Representations and Mixture-of-Experts Policy for Robust Multi-Task Robotic Manipulation
• Fast Visuomotor Policy for Robotic Manipulation
• Failure Prediction at Runtime for Generative Robot Policies
• VER: Vision Expert Transformer for Robot Learning via Foundation Distillation and Dynamic Routing
• Policy Compatible Skill Incremental Learning via Lazy Learning Interface
• From Observation to Insight: Mechanistic World Models and the Quest for Autonomous Discovery
• Robotic Manipulation via Imitation Learning: Taxonomy, Evolution, Benchmark, and Challenges
• Integrating Attention-Enhanced LSTM and Particle Swarm Optimization for Dynamic Pricing and Replenishment Strategies in Fresh Food Supermarkets
• AutoRing: Imitation Learning--based Autonomous Intraocular Foreign Body Removal Manipulation with Eye Surgical Robot
• Grounding Actions in Camera Space: Observation-Centric Vision-Language-Action Policy
• ADPro: a Test-time Adaptive Diffusion Policy via Manifold-constrained Denoising and Task-aware Initialization for Robotic Manipulation
• Learning to See and Act: Task-Aware Virtual View Exploration for Robotic Manipulation
• VFP: Variational Flow-Matching Policy for Multi-Modal Robot Manipulation

Related

• Flexible Multitask Learning with Factorized Diffusion Policy
• Diffusion Policy: Visuomotor Policy Learning via Action Diffusion
• Diffusion policy: Visuomotor policy learning via action diffusion
• DiWA: Diffusion Policy Adaptation with World Models
• Diffusion Policy Policy Optimization