Keeping it Real: Using Artificial Intelligence to Deliver Practical Outcomes for Road Managers

Abstract

In a world of ever-smarter technologies, how do we as Road Managers avoid getting caught up in “tech hype” and successfully navigate endless new possibilities to deliver real outcomes in the way we manage our community’s roads?

This paper aims to answer that question by sharing the lessons learned by the Regional Road Managers Working Group who have been progressing the award-winning 2021 Roads Alliance Project – Using Artificial Intelligence for Road Maintenance Management.

The advancement of the Roads Alliance Project in the last 12 months has included: - Improving machine learning confidence; and - Significantly improving the method of displaying the AI data as well as the speed and functionality of using the data to support decision making.

Conversely some of the challenges have included: - The huge volume of data AI returns; and - Bias in data that can fatally compromise generalisation and real-world performance.

Case studies from the Working Group demonstrate the evolution of AI road management technology and how building a solution with the end-user in mind e.g. the under-resourced and time poor Regional Road Manager has kept it real.

Key lessons learned include: - Preventing the risk of getting lost in the technology by starting with a clearly defined and narrow focus i.e. High Priority Safety Defects; - Computers are not perfect – the human value and relationship with data e.g. human triage and ground-truthing; - The importance of starting with a system that is already working and understanding where and how technology will make real gains; - The user experience – using technology to focus on what’s important while the human remains the decision maker; and - The risk of relying on technology to identify defects without using our expertise as Road Managers to understand the root cause, therefore we are no longer effectively managing the asset.

It has been a fascinating journey so far and we look forward to sharing the struggles and demonstrating the successes of using AI to deliver practical outcomes to regional road managers and better roads for local communities