# Cross-Modality Alignment Perception and Multi-Head Self-Attention Mechanism for Vision-Language-Action of Humanoid Robot

## TL;DR

The paper introduces a new attention mechanism and training strategy for humanoid robots to improve task performance and reduce computational load.

## Contribution

A memory-gated filtering attention model and cross-modal alignment perception with few-shot data collection for Vision-Language-Action tasks in humanoid robots.

## Key findings

- The proposed model reduced video memory usage by 72% and improved training speed from 1.35 s to 0.129 s per batch.
- The Vision-Language-Action system improved task success rate and reduced robot arm jitter during complex operations.

## Abstract

What are the main findings?
A model of memory-gated filtering attention was proposed, which improved multi-head self-attention mechanism.A cross-modal alignment perception during training was designed, which combined with a few-shot data collection strategy of key steps.

A model of memory-gated filtering attention was proposed, which improved multi-head self-attention mechanism.

A cross-modal alignment perception during training was designed, which combined with a few-shot data collection strategy of key steps.

What are the implications of the main findings?
The proposed multi-head self-attention mechanism could reduce the video memory occupation by 72% and improve the training speed from 1.35 s to 0.129 s per batch.The proposed Vision-Language-Action of a humanoid robot significantly improved the task success rate and alleviated the robot arm jitter problem.

The proposed multi-head self-attention mechanism could reduce the video memory occupation by 72% and improve the training speed from 1.35 s to 0.129 s per batch.

The proposed Vision-Language-Action of a humanoid robot significantly improved the task success rate and alleviated the robot arm jitter problem.

For a humanoid robot, it is difficult to predict a motion trajectory through end-to-end imitation learning when performing complex operations and multi-step processes, leading to jittering in the robot arm. To alleviate this problem and reduce the computational complexity of the self-attention module in Vision-Language-Action (VLA) operations, we proposed a memory-gated filtering attention model that improved the multi-head self-attention mechanism. Then, we designed a cross-modal alignment perception during training, combined with a few-shot data-collection strategy for key steps. The experimental results showed that the proposed scheme significantly improved the task success rate and alleviated the robot arm jitter problem, while reducing video memory usage by 72% and improving training speed from 1.35 s to 0.129 s per batch. This maintained higher action accuracy and robustness in the humanoid robot.