Smart goggle for personal mobility

Abstract

Overtaking prediction system for smart vehicles has always been a famous research topic worldwide, by taking into considerations the driver’s gaze information as well as the data collected by sensors installed inside the vehicles. In Singapore, Personal Mobility Devices (PMDs) are often the choice of transportation for its citizens, as it can help to reduce the total emission of greenhouse gases in the “Clean City”. Unfortunately, the usage of PMDs in Singapore is often related with accidents and injuries. Therefore, this project aims to develop an “Overtaking Prediction System” for PMDs, using both rider’s gaze information and his eye view (camera) as input. Using a pre-trained YOLO Machine Learning model and a self-developed Overtaking Prediction System, the rider’s overtaking intention can then be used to turn on a “Turn Left” or a “Turn Right” signalling light on PMD, without requiring rider’s action. Doing this can help to prevent collision between a fast-moving or reckless rider and other people around him. Real-time experiments involving 7 participants of ages ranging from 23 to 31 years old, or a total of 234 overtaking events, are carried out to examine the accuracy of the smart system, and the result is shown as 84.62%. This shows that PMD rider’s gaze data can be a good source of input to determine his overtaking intention. This project also serves as the first step to design Advanced Driver Assistance Systems (ADAS) for PMDs, which have not existed in the market currently.