Collision Classification for Unmanned Surface Vehicle using Inertial Measurement Unit Data

In the development of digital twin systems for unmanned surface vehicles (USVs), collision classification is a fundamental feature to implement. Integrating collision position information into remote control systems significantly enhances operator situational awareness and facilitates effective motion control strategies even when camera visitor is compromised by challenging environmental conditions. This paper presents a three-step approach for collision classification employing onboard inertial measurement unit (IMU) data. In the first step, the sensor measures proprioceptive data during the emulated collision between the remote-controlled model scale ship and a stationary obstacle. Secondly, this time-series data is processed via wavelet transform followed prior to being fed into a deep neural network in the third step to classify collision into three types: bow-side (front), port-side (left), and starboard-side (right) collisions. The achieved classification accuracy of 96.66% shows the promising potential for integrating this approach into real-time digital twin systems.