Smarter Roads.
Faster Ambulances. Safer Cities.
An ML-powered traffic ecosystem that dynamically controls city signals, prioritizes emergency vehicles, detects potholes, and flags traffic violations — built and won at Smart India Hackathon 2019 against teams from across the entire country.
India's traffic congestion problem isn't just inconvenient — it's lethal. Every day, 1,214 road crashes occur across the country. Most are caused by traffic violations or congestion-induced risk. In 2017 alone, 3,597 people were killed due to potholes.
Intelligent Transport Systems (ITS) exist in developed nations, but the cost of implementation, maintenance, and per-vehicle tracking is prohibitively high for India. Emergency vehicles — ambulances, fire engines — are stuck in the same traffic as everyone else. Not a single second saved when lives are on the line.
The Maruti Suzuki problem statement at Smart India Hackathon 2019 challenged teams across India: build a cost-effective, scalable, ML-driven traffic system that could actually be deployed on Indian roads — without replacing existing infrastructure.
Instead of tracking every vehicle — computationally expensive and data-heavy — we use a Bay Detection approach: horizontal slot words painted on road surfaces are monitored by existing CCTV cameras. When vehicles obscure the words, density is measured instantly using OCR. This eliminates costly per-vehicle sensors entirely.
The Smart Traffic Controller (STC) device installs on any existing traffic light pole, making the system fully backward-compatible. It receives density values from the cloud and triggers signals dynamically — no fixed timing, no manual control required.
For emergency vehicles, our Intelligent Navigation System (INS) syncs with STC automatically when an ambulance enters junction range, creating an instant green corridor. Regular drivers are simultaneously notified via voice alerts. The system is self-healing: if the network drops, STC falls back to SMPP-based SMS, maintaining zero downtime.
Instead of tracking every vehicle, we painted horizontal slot words (BAY 1, BAY 2, BAY 3) on road surfaces. CCTV cameras use Tesseract OCR + OpenCV to detect which words are hidden by vehicles — giving a real-time density level (0–3) that dynamically triggers signal timing. No fixed timers, no wasted green time.
A dedicated GPS-based navigation system for emergency and regular vehicles. As an ambulance approaches a junction, INS syncs wirelessly with STC, turns that lane green, forces all others red — with zero driver interaction. Nearby drivers simultaneously receive: 'Save a Life! Please keep left as ambulance is heading forward.'
When a signal is red, the system scans for the stop line using computer vision. Any vehicle crossing the line is flagged — its license plate extracted via OCR and stored in the cloud with a timestamp. If CCTV is absent and the driver uses INS, the app captures the violation and pushes it to the server automatically.
A trained object detection model running on dashboard cameras identifies potholes in real time with up to 99% confidence. The exact GPS coordinates are sent to the server and plotted on INS maps. Drivers approaching that location receive a geofenced alert — crowdsourcing road safety data for proactive government maintenance.
The system scrapes news articles and web headlines to detect protests, parades, VIP convoys, and accidents. Locations are extracted from the JSON payload and plotted as markers on the INS map. Data refreshes continuously so drivers re-route before reaching congestion. Police can also lock down road segments remotely.
A centralized dashboard gives traffic police remote control over every signal in the city. Operators can view real-time density data, change signal states, block roads for special events, view live CCTV feeds, and monitor violation logs — all without stepping onto the street.
The STC is a hardware device installed on existing traffic light poles — no infrastructure replacement required. It runs on 240V AC converted to 12V DC, with a 4-channel relay. It communicates via a private network with the server. When offline, it auto-falls back to SMPP-encoded SMS via Non-cellular Data Transmission, ensuring zero downtime.
GPS modules embedded in ambulances broadcast real-time location. When the vehicle enters the pre-defined radius of a junction, INS and STC auto-sync: that lane turns green, all others red, and the state pushes to cloud — alerting all drivers along the route. The system coordinates multiple junctions simultaneously along the full emergency corridor.
Each lane is divided into 3 bays with words painted on the road. CCTV cameras with OCR check which labels are hidden by vehicles, generating a density level 0–3. The STC device fires the exact green duration — no vehicles, no wasted green time.
All bays visible. No vehicles. Green not triggered — saves unnecessary wait cycles.
BAY 1 hidden. Low density. 20 seconds of green — sufficient for queue clearance.
BAY 1 & 2 hidden. Medium density. 40 seconds for moderate traffic throughput.
All 3 bays hidden. Max density. Full 60-second green for complete intersection clearance.
Registered under the Ministry of Commerce and Industry, Government of India. Application number: 202041024036. The entire concept — Bay Detection algorithm and STC hardware design — is IP-protected.
View Patent Document →Peer-reviewed research published in Springer Journal. Documents the full system architecture, ML models, and deployment results from the Smart India Hackathon 2019 prototype.
DOI: 10.1007/978-3-030-77637-4_2 →Official software concept copyright registration covering the intelligent Bay Detection System and STC hardware interactions.
View Copyright Document →
Grand Finale — Software Edition, 2nd & 3rd March 2019 at Banaras Hindu University, Varanasi. Team Musketeers was officially declared #1 by Maruti Suzuki and the Government of India.
Telecasted and published across multiple news channels and newspapers in India.
View media coverage →Recognized by MHRD (Ministry of Human Resource Development), Government of India, and endorsed by Maruti Suzuki at the national level.
Team met with Prof. Anil D. Sahasrabudhe, Chairman of All India Council for Technical Education, at the Grand Finale ceremony.
Deputy General Manager Maruti Suzuki India Limited
Senior Manager Maruti Suzuki India Limited
A team of 7 engineers from Panimalar Engineering College, Chennai built this system end-to-end — hardware, ML, mobile, and cloud — in under 36 hours at the Grand Finale.
Sundeep Dayalan served as Lead Developer, architecting the full system and coordinating every module. The team name — Musketeers — is signed on the hardware prototype itself.
The project spanned from January 2019 to March 2019, covering ideation, prototype, hackathon win, patent filing, media appearances, and final Springer publication.