Two-layer autonomous intelligence dynamic trajectory planning method based on shield-tunnel ring-geology interactions

Abstract

To solve the problem that current attitude planning methods do not fully consider the interaction and constraints among the shield, segmental tunnel ring, and geology, and cannot adapt to the changes in the actual engineering environment, or provide feasible long-term and short-term attitude planning, this paper proposes autonomous intelligent dynamic trajectory planning (AI-DTP) to provide tunnel ring and centimeter-layer planning targets for a self-driving shield to meet long-term accuracy and short-term rapidity. AI-DTP introduces the Frenet coordinate system to solve the problem of inconsistent spatial representation of tunnel data, segmental tunnel ring location, and surrounding geological conditions, designs the long short-term memory attitude prediction model to accurately predict shield attitude change trend based on shield, tunnel, and geology, and uses a heuristic algorithm for trajectory optimization. AI-DTP provides ring-layer and centimeter-layer planning objectives that meet the needs of long-term accuracy and short-term correction of shield attitude control. In the Hangzhou-Shaoxing Intercity Railroad Tunnel Project in China, the “Zhiyu” shield equipped with the AI-DTP system was faster and more accurate than the manually controlled shield, with a smoother process and better quality of the completed tunnel.

Keywords

• Shield tunnel
• Trajectory planning
• Attitude control
• Self-driving
• Machine learning