New Delhi, Sept. 21 -- The government is testing an artificial intelligence system that can identify potholes, cracks and other defects on rural roads from videos recorded on mobile phones, with a nationwide field-validation exercise underway since September 1, a senior official said on Sunday. The ministry of rural development has given "in-principle approval" for trials of the AI-based road-condition assessment system on roads built under the Pradhan Mantri Gram Sadak Yojana (PMGSY). The system is being developed by the Centre for Development of Advanced Computing with the National Rural Infrastructure Development Agency. At present, road maintenance is monitored through physical inspections and geo-tagged photographs recorded by field officials on the eMARG mobile application. According to a senior government official, the technology is only being tested as a tool to support road maintenance decisions. Engineers will continue to physically verify road defects, comparing their measurements with the findings generated by the AI system, the official said. "The AI-based road condition assessment is being developed and field-validated to support efficient and technology-enabled road maintenance assessment". The initial trials covered six roads in Pune, Lucknow district, Kamrup district in Assam and Ri-Bhoi district in Meghalaya. Trials using an improved version of the model were subsequently conducted on seven roads in Kanpur district and Berasia block of Bhopal district, as per the official cited above. "The system is being tested to identify visible road defects including potholes, longitudinal and transverse cracks, edge breaks, surface depressions, patches and vegetation-related obstructions," he added. Engineers however, will physically measure these defects and compare them with the AI output. "The exercise is intended to provide more data to refine the system before wider use". A wider national field-validation exercise began on September 1, with states and Union territories identifying Programme Implementation Units (PIUs) for trials on PMGSY roads. Currently, a mobile application is being used to record road-condition videos to check their quality, using a mobile device already mounted on vehicles available with PIUs. The government said dedicated dashcams could be considered later depending on field requirements and operational suitability. "The approach is meant to allow road-condition assessments without creating a separate fleet of inspection vehicles," the official said. The ministry is also considering whether roads could eventually be assessed every quarter under a proposed technology-enabled maintenance framework. However, the frequency has not yet been finalised. The ministry also said existing contractual provisions, including those relating to maintenance and the Defects Liability Period, will continue to apply. "If an AI assessment differs from an engineer's assessment, the difference will be examined during the field-validation exercise. A verification and review protocol will be finalised before wider implementation," the ministry explained, responding to HT questions. "Wider deployment will depend on the outcome of the national validation exercise, operational readiness and further approvals," it said....