Project Summary
his proposal for the first time develops a machine learning model that can identify potholes by learning from the behaviours of the riders, both when a cyclist rides over a pothole and when a cyclist manages to avoid a pothole.
Project Achievements
A mobile application and a machine learning algorithm have been developed for the purpose of collecting sensor data. By conducting a number of experiments of different types of riding behaviour, we have collected the sensor data for recognising the pattern when encountering a pothole. The key technology is a new algorithm utilising machine learning to study the pattern of riding around a pothole.
Conclusions
The data including accelerometer, gyroscope and GPS have been collected by the app as a prototype. The machine learning model for pattern recognition delivers some good initial results by considering one dimension in gyropscope.
The project has hit several key milestones in the plan and shown the potential of improving riding experience and the infrastructure. However, the technology is still at proof-ofconcept stage. With further work, the algorithm can be improved and more factors can be taken into account in a wider programme.
