AI Technology Detects More Than 300 Road Defects in Victorian Trial

Melbourne — An artificial intelligence-powered monitoring system has detected more than 300 road defects across 1,000 kilometres of Victorian roads during a two-month trial, as researchers explore ways to help authorities identify deterioration earlier and better target maintenance.

Developed by the Royal Melbourne Institute of Technology (RMIT) University with industry partners, and tested in collaboration with Royal Automobile Club of Victoria (RACV), the RoughSense system was installed in an emergency roadside assistance van operating six days a week between Balnarring and Dromana on the Mornington Peninsula.

The trial ran from 21 August to 21 October, with defects most frequently recorded along the Frankston-Flinders Road corridor through Tyabb, Hastings, Bittern and Balnarring.

RoughSense uses artificial intelligence, GPS tracking, video footage, sensors and vehicle movement data to detect potholes, cracks, rutting and early signs of road deterioration as the vehicle follows its normal routes.

Its findings are processed in near real time, while the driver manually flags defects to help validate the system’s detections.

RMIT Project Lead, Dr. Amirali Khodadadian Gostar, said the RACV vehicle provided a practical means of evaluating the technology under everyday road conditions.

“The RACV van was a critical mechanism to test the effectiveness of our defect detecting device,” he said.

Associate Professor, Hormoz Marzbani, an RMIT Co-Investigator, said repeated journeys had identified potholes that were getting worse, demonstrating how ongoing monitoring could reveal changes that a single inspection might miss.

The research aims to help road authorities move beyond repairs prompted by visible damage or complaints towards maintenance informed by regularly collected road-condition data.

The trial has demonstrated the system's ability to record defects, although its effectiveness at a wider scale and its impact on maintenance costs remain to be established.

RACV Head of Public Policy, James Williams, said the project addressed concerns raised by motorists in the organisation’s 2026 My Country Road survey about potholes and deteriorating roads.

"This trial demonstrates how technology can provide valuable insights into road conditions, helping support more targeted maintenance planning and investment decisions,” Williams said.

RACV facilitated and supported the trial, which is funded through the National Road Action Grant.

Mornington-based patrol driver, Mark Orlando, operates the vehicle and helps researchers verify detections.

Orlando said some dangerous defects were difficult for motorists to see, particularly in wet weather, at night or on unfamiliar roads.

“Anything that helps identify hazardous road conditions sooner can improve safety and reduce vehicle damage,” he said.

RMIT leads the research programme, system integration and development of sensor-data fusion and AI-based defect detection.

Industry partner MetaSpatial contributes sensing hardware, system implementation, field testing and data visualisation.

MetaSpatial Director, Dr. Adam Chabok said the project illustrated the potential of combining university research with industry expertise to address road safety challenges.

The trial will continue until January 2027, with researchers working to improve the system’s accuracy, reliability and ability to track changes in road conditions over time.

The next phase will examine whether it can be deployed across larger vehicle fleets to provide more extensive, continuous monitoring and support preventative maintenance across Australia.