BENGALURU: Three Artificial Intelligence-based solutions aimed at predicting traffic congestion, improving deployment of traffic police personnel and tackling parking related bottlenecks have emerged as winners of Gridlock Hackathon 2.0, organised by Flipkart in collaboration with Bengaluru Traffic Police (BTP).
The hackathon attracted 33,600 participants from across India and resulted in more than 1,100 AI and machine-learning prototypes developed using anonymised, real-world Bengaluru traffic data. The first-place solution, Gridlock Oracle, uses a Hawkes Process model to predict how a single traffic incident could trigger cascading congestion across the city. It also provides a live control-room dashboard intended to help traffic police deploy resources before congestion spreads.
PRAHAR (Predictive Resource Allocation for High-Impact Area Response), which secured second place, uses multiple machine-learning models to forecast congestion severity, junction risks, incident resolution time and cascading effects. The platform also recommends deployment of personnel, barricades, diversion routes and dispatch stations.
Third-place winner ParkSight -- Parking-Induced Congestion Intelligence -- analysed parking violation records to identify congestion and enforcement hotspots and proposed measures, including forecasting and patrol optimisation.
The hackathon was conducted in association with MapmyIndia, Arcadis and HackerEarth. Participants used datasets made available through the BTP’s ASTraM (Actionable Intelligence for Sustainable Traffic Management) platform. The finale was held at Flipkart’s Bengaluru headquarters in the presence of Bengaluru City Police Commissioner Seemant Kumar Singh.
The second edition follows the inaugural 2017 hackathon, whose winning traffic-signal solution was tested at Silk Board Junction and indicated a potential reduction of about 17 per cent in waiting time.