Scholarlink

This paper introduces a novel terrain recognition method designed for unstructured environments, combining BiLSTM and Transformer encoder networks. The BiLSTM effectively captures temporal dependencies in the data, while the Transformer’s attention mechanism models the correlations between control commands and trajectory information. The proposed method demonstrates an impressive accuracy of 0.9696 on the test set. Furthermore, a deployment strategy is outlined for integrating the method within the modern multi-domain electronic and electrical (E/E) architecture of unmanned ground vehicles, providing valuable insights for real-world applications.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Terrain Recognition for Unstructured Environments Using BiLSTM-Transformer Networks Under Multi-domain E/E Architecture

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jie Fan,
  • Xudong Zhang,
  • Yijie Chen,
  • Jiangbo Geng,
  • Yutong Jiang,
  • Xuan Liu

摘要

This paper introduces a novel terrain recognition method designed for unstructured environments, combining BiLSTM and Transformer encoder networks. The BiLSTM effectively captures temporal dependencies in the data, while the Transformer’s attention mechanism models the correlations between control commands and trajectory information. The proposed method demonstrates an impressive accuracy of 0.9696 on the test set. Furthermore, a deployment strategy is outlined for integrating the method within the modern multi-domain electronic and electrical (E/E) architecture of unmanned ground vehicles, providing valuable insights for real-world applications.