A bionic underwater robot body perception autonomous obstacle avoidance method based on a double-path 1D-CNN
By integrating a flexible sensor array and a dual-path 1D-CNN onto a biomimetic underwater robot, efficient identification and obstacle avoidance control of collision events in complex underwater environments are achieved, solving the problems of response delay and large blind spots in existing technologies and improving the obstacle avoidance capabilities of underwater robots.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NINGBO INST OF NORTHWESTERN POLYTECHNICAL UNIV
- Filing Date
- 2026-02-05
- Publication Date
- 2026-06-02
AI Technical Summary
Existing underwater obstacle avoidance technologies suffer from large blind spots, response delays, and sensitivity to light and water quality in complex aquatic environments. They are unable to achieve highly sensitive, robust recognition and timely response to obstacles at close range, making biomimetic underwater robots prone to structural damage or mission failure.
A biomimetic underwater robot body perception method based on dual-path 1D-CNN is adopted. The robot collects motion signals of the body structure through a flexible sensor array, extracts features using time-domain and frequency-domain branch networks, and combines feature fusion and collision discrimination layers to predict collision probability and issue obstacle avoidance control commands to adjust the robot's motion posture.
It achieves highly sensitive and robust recognition and timely response to collision events in complex underwater environments, reduces false alarms and missed alarms, improves the real-time performance and reliability of obstacle avoidance decisions, and ensures the safe movement of robots.
Smart Images

Figure CN122131795A_ABST