Project Summary

This project seeks to revolutionise the reporting of road conditions for enhanced safety during travel and efficient maintenance by local authorities. The team will build a hardware stack that monitors journey activity and communicates with local authorities. The cloud-based smart bike incorporates an onboard accelerometer sensor and embedded machine learning to capture data about road surface quality from the perspective of cyclists. By providing this valuable information, the solution aims to optimise resource allocation and streamline labor requirements, thus saving time and money for local authorities. Ultimately, the project aims to prioritise and allocate resources where they are needed most for better road safety and maintenance.

Project Achievements

The project was divided into three main workstreams: 1. Exploration – We identified viable hardware and software tools. 2. Data collection – We identified the events we wanted to detect, a mix of continuous events (road surface quality), and discrete events (potholes, hard braking, speed bumps). We then collected the data using smartphones with the Sensor Logger application fitted to several bicycles. 3. Machine learning – We classified events, and built a machine learning model using a convolutional neural network. The model classifies events at 87.8% accuracy and we tested this using raw data. The results can be seen in the video here.

Conclusions

• The project successfully demonstrated our ability to use data from bicycles to identify road surface quality and rider events. We successfully moved the project into TRL4. • We were not able to integrate the hardware directly into the bicycle, but we have furthered our understanding of the existing hardware capabilities, and hardware requirements for a future prototype. • Being part of the TRIG programme gave us the opportunity to explore this idea further and dedicate resources to do a technical research project alongside developing the business.

Next Steps

• Increasing our data sample size with an expanded test programme deploying multiple riders. • And using this to increase accuracy levels to a target of 95%. • Expand the scope of road surface qualities and rider behaviours we are able to capture. • Consulting with potential collaborators (e.g. fleet operators, local councils) to understand effective ways of deploying our technology for the benefit of end users). • Securing follow-on funding to progress beyond TRL4.