NHAI deploys in-house AI system to identify defects in highway project reports and road infrastructure
The National Highway Authority of India (NHAI) has developed an in-house artificial intelligence system to detect deficiencies in highway project reports and road defects. The system, recently deployed, uses NHAI's repository of documents including circulars, acts, Indian Roads Congress (IRC) codes, and monthly reports.
The AI tool assists NHAI officials by reducing the need to manually review hundreds of pages of guidelines and standard operating procedures. Currently, only NHAI officials have access to the system. NHAI Chairman Santosh Kumar Yadav said access will be extended to consultants and contractors once safeguards against misuse are in place.
Yadav explained that the system was developed after a workflow study revealed that project directors and regional officers were spending time on tasks that could be automated. Officials can now upload a photo of a highway defect, and the system checks it against guidelines and codes to provide results.
The system also addresses issues in Detailed Project Reports (DPRs), which Union Minister for Road Transport and Highways Nitin Gadkari has cited as a factor in road accidents and fatalities. By uploading a DPR, the AI can identify errors against IRC codes and policy circulars, allowing corrections in advance.
NHAI's AI team also created a Technical Schedule Analyzer tool that reviews Schedule B and C documents of DPRs, flagging issues by severity and completeness. Additionally, a chat-based AI agent called Margsarthi allows officials to ask queries. For example, a field engineer can photograph a problem and ask Margsarthi for guidance. The tool, live since April 18, has received over 50,000 queries from about 1,100 users, with over 40% related to circulars and documents.
Other tools include Kick-Off and MPR insights, which track issues from kick-off meetings and monthly project reports, flag critical bottlenecks, and provide recommended actions with source citations.