DRAE: Dynamic Retrieval-Augmented Expert Networks for Lifelong Learning and Task Adaptation in Robotics

2025/07/07 by Long, Yayu, Chen, Kewei, Jin, Long +1 · 4 citations
#FOS: Computer and information sciences #I.2.6 #I.2.9 #Robotics (cs.RO)

paper · doi:10.48550/arxiv.2507.04661

Abstract

We introduce Dynamic Retrieval-Augmented Expert Networks (DRAE), a groundbreaking architecture that addresses the challenges of lifelong learning, catastrophic forgetting, and task adaptation by combining the dynamic routing capabilities of Mixture-of-Experts (MoE); leveraging the knowledge-enhancement power of Retrieval-Augmented Generation (RAG); incorporating a novel hierarchical reinforcement learning (RL) framework; and coordinating through ReflexNet-SchemaPlanner-HyperOptima (RSHO).DRAE dynamically routes expert models via a sparse MoE gating mechanism, enabling efficient resource allocation while leveraging external knowledge through parametric retrieval (P-RAG) to augment the learning process. We propose a new RL framework with ReflexNet for low-level task execution, SchemaPlanner for symbolic reasoning, and HyperOptima for long-term context modeling, ensuring continuous adaptation and memory retention. Experimental results show that DRAE significantly outperforms baseline approaches in long-term task retention and knowledge reuse, achieving an average task success rate of 82.5% across a set of dynamic robotic manipulation tasks, compared to 74.2% for traditional MoE models. Furthermore, DRAE maintains an extremely low forgetting rate, outperforming state-of-the-art methods in catastrophic forgetting mitigation. These results demonstrate the effectiveness of our approach in enabling flexible, scalable, and efficient lifelong learning for robotics.

Citations

• Learning from Massive Human Videos for Universal Humanoid Pose Control
• T2M-X: Learning Expressive Text-to-Motion Generation from Partially Annotated Data
• HybridRAG: Integrating Knowledge Graphs and Vector Retrieval Augmented Generation for Efficient Information Extraction
• DRAMA: An Efficient End-to-end Motion Planner for Autonomous Driving with Mamba
• Sparse Diffusion Policy: A Sparse, Reusable, and Flexible Policy for Robot Learning
• Human-Aware Vision-and-Language Navigation: Bridging Simulation to Reality with Dynamic Human Interactions
• NAVSIM: Data-Driven Non-Reactive Autonomous Vehicle Simulation and Benchmarking
• Octo: An Open-Source Generalist Robot Policy
• Learning More Generalized Experts by Merging Experts in Mixture-of-Experts
• From Local to Global: A Graph RAG Approach to Query-Focused Summarization
• SkillDiffuser: Interpretable Hierarchical Planning via Skill Abstractions in Diffusion-Based Task Execution
• Catastrophic Forgetting in Deep Learning: A Comprehensive Taxonomy
• SPOC: Imitating Shortest Paths in Simulation Enables Effective Navigation and Manipulation in the Real World
• CareerX: A Retrieval-Augmented Generation Framework for Personalized AI-Driven Career Guidance
• Enhancing NeRF akin to Enhancing LLMs: Generalizable NeRF Transformer with Mixture-of-View-Experts
• LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning
• DexArt: Benchmarking Generalizable Dexterous Manipulation with Articulated Objects
• MotionDiffuse: Text-Driven Human Motion Generation with Diffusion Model
• TransFuser: Imitation with Transformer-Based Sensor Fusion for Autonomous Driving
• Lifelong Robotic Reinforcement Learning by Retaining Experiences
• LoRA: Low-Rank Adaptation of Large Language Models
• MVSNeRF: Fast Generalizable Radiance Field Reconstruction from Multi-View Stereo
• IBRNet: Learning Multi-View Image-Based Rendering
• pixelNeRF: Neural Radiance Fields from One or Few Images
• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
• A tutorial on Dirichlet process mixture modeling
• Memory Aware Synapses: Learning what (not) to forget
• Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
• FeUdal Networks for Hierarchical Reinforcement Learning
• Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer
• The Option-Critic Architecture
• Progressive Neural Networks

Cited by

• Robust Finetuning of Vision-Language-Action Robot Policies via Parameter Merging
• FT-NCFM: An Influence-Aware Data Distillation Framework for Efficient VLA Models
• Dynamic Nested Hierarchies: Pioneering Self-Evolution in Machine Learning Architectures for Lifelong Intelligence
• ExpReS-VLA: Specializing Vision-Language-Action Models Through Experience Replay and Retrieval

Related

• Dynamic Mixture of Progressive Parameter-Efficient Expert Library for Lifelong Robot Learning
• Task-agnostic Lifelong Robot Learning with Retrieval-based Weighted Local Adaptation
• AURA: Autonomous Upskilling with Retrieval-Augmented Agents
• Towards General Purpose Robots at Scale: Lifelong Learning and Learning to Use Memory
• LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning