Flipkart and the Bengaluru Traffic Police have brought together developers and technology professionals to explore artificial intelligence and machine learning-based solutions for Bengaluru's traffic congestion, with more than 1,100 prototypes developed through the second edition of the Gridlock Hackathon.
The Gridlock Hackathon 2.0 attracted 33,600 registrations from students, developers and working professionals across India. Conducted in collaboration with MapmyIndia, Arcadis and HackerEarth, the initiative focused on using technology to address congestion and broader urban mobility challenges in Bengaluru.
Participants worked with anonymised, real-world traffic datasets made available through the Bengaluru Traffic Police's ASTraM, or Actionable Intelligence for Sustainable Traffic Management, platform, which was designed and implemented by Arcadis.
The datasets were used to develop AI and machine learning models across areas including congestion prediction, traffic violation detection, movement-pattern analysis, parking intelligence and mobility decision-making. The challenge progressed from an online machine learning stage to prototype development for shortlisted participants.
Three solutions were selected as winners.
The first-place solution, Gridlock Oracle, uses a Hawkes Process model to predict how an individual traffic incident could trigger congestion across other parts of the city. The system includes a control-room dashboard designed to help traffic authorities anticipate congestion and deploy resources before it spreads.
PRAHAR, which placed second, uses multiple machine learning models to forecast congestion severity, junction risks, resolution times and possible cascading effects. It is also designed to recommend the deployment of traffic officers, barricades, diversion routes and dispatch stations.
Third-placed ParkSight focuses on parking-related congestion. The system analyses parking violation data to identify priority congestion and enforcement areas while incorporating hotspot detection, forecasting and patrol optimisation.
The three teams shared a prize pool of Rs 5 lakh, with the winning team receiving Rs 2.25 lakh, followed by Rs 1.75 lakh and Rs 1 lakh for the second and third-place teams, respectively.
Gridlock 2.0 follows Flipkart's first traffic-focused hackathon in 2017, which attracted close to 3,000 registrations. A solution from that edition proposed using real-time traffic information for dynamic traffic-signal switching, with testing at Bengaluru's Silk Board Junction indicating a potential reduction of about 17% in waiting time.
The latest edition reflects the growing use of AI and real-world datasets to explore urban mobility problems. However, Flipkart and the Bengaluru Traffic Police have not announced that the three winning systems will be deployed across the city's traffic network.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.