Few-Shot Vision-Language Action-Incremental Policy Learning

2025-04-22

AI Summary

This work addresses catastrophic forgetting in robotic manipulation under few-shot and continually emerging task settings. To this end, we propose Few-Shot Action Incremental Learning (FSAIL), a novel paradigm for incremental skill acquisition. Methodologically, we establish the first few-shot continual learning framework tailored for action skills, introducing a task prompt graph evolution mechanism that fuses text–vision multimodal cues. Our approach incorporates a Transformer-based multi-view spatial representation module, task-specific prompt (TSP) learning, continuous evolution strategy (CES), and a multimodal deep interaction module to enable cross-task skill reuse. Evaluated on standard benchmarks, FSAIL achieves over 26% improvement in success rate, significantly enhancing the few-shot adaptability and continual learning robustness of vision-language policies.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multimodal Learning
• Multiagent Systems: Multiagent Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Recently, Transformer-based robotic manipulation methods utilize multi-view spatial representations and language instructions to learn robot motion trajectories by leveraging numerous robot demonstrations. However, the collection of robot data is extremely challenging, and existing methods lack the capability for continuous learning on new tasks with only a few demonstrations. In this paper, we formulate these challenges as the Few-Shot Action-Incremental Learning (FSAIL) task, and accordingly design a Task-prOmpt graPh evolutIon poliCy (TOPIC) to address these issues. Specifically, to address the data scarcity issue in robotic imitation learning, TOPIC learns Task-Specific Prompts (TSP) through the deep interaction of multi-modal information within few-shot demonstrations, thereby effectively extracting the task-specific discriminative information. On the other hand, to enhance the capability for continual learning on new tasks and mitigate the issue of catastrophic forgetting, TOPIC adopts a Continuous Evolution Strategy (CES). CES leverages the intrinsic relationships between tasks to construct a task relation graph, which effectively facilitates the adaptation of new tasks by reusing skills learned from previous tasks. TOPIC pioneers few-shot continual learning in the robotic manipulation task, and extensive experimental results demonstrate that TOPIC outperforms state-of-the-art baselines by over 26$%$ in success rate, significantly enhancing the continual learning capabilities of existing Transformer-based policies.

Problem

• Addresses few-shot learning for robot tasks with minimal demonstrations
• Enhances continual learning to prevent forgetting previous task skills
• Improves robotic manipulation using multi-modal task-specific prompts

Innovation

• Task-Specific Prompts for few-shot learning
• Continuous Evolution Strategy for task adaptation
• Task relation graph for skill reuse