CAGE: Causal Attention Enables Data-Efficient Generalizable Robotic Manipulation

TL;DR

CAGE introduces a causal attention-based robotic manipulation policy that, with minimal demonstrations, generalizes effectively across diverse environments, outperforming existing methods in real-world manipulation tasks.

Contribution

The paper presents CAGE, a novel manipulation policy integrating causal attention, vision foundation models, and diffusion-based action prediction for improved generalization with limited data.

Findings

Achieves 42% increase in task completion rate over state-of-the-art methods.

Attains 43% completion and 51% success rates in unseen environments.

Outperforms existing RGB/RGB-D approaches under large distribution shifts.

Abstract

Generalization in robotic manipulation remains a critical challenge, particularly when scaling to new environments with limited demonstrations. This paper introduces CAGE, a novel robotic manipulation policy designed to overcome these generalization barriers by integrating a causal attention mechanism. CAGE utilizes the powerful feature extraction capabilities of the vision foundation model DINOv2, combined with LoRA fine-tuning for robust environment understanding. The policy further employs a causal Perceiver for effective token compression and a diffusion-based action prediction head with attention mechanisms to enhance task-specific fine-grained conditioning. With as few as 50 demonstrations from a single training environment, CAGE achieves robust generalization across diverse visual changes in objects, backgrounds, and viewpoints. Extensive experiments validate that CAGE significantly outperforms existing state-of-the-art RGB/RGB-D approaches in various manipulation tasks, especially under large distribution shifts. In similar environments, CAGE offers an average of 42% increase in task completion rate. While all baselines fail to execute the task in unseen environments, CAGE manages to obtain a 43% completion rate and a 51% success rate in average, making a huge step towards practical deployment of robots in real-world settings. Project website: cage-policy.github.io.