AI-Powered Mobile Urban Intelligence Platform Using Public Transport Fleet
“City buses have cameras, but they only record video for later viewing instead of helping fix roads or manage traffic. This solution turns standard public buses into smart mobile scanners that automatically detect potholes, traffic jams, and reckless drivers while driving, sending live updates to a central dashboard for city authorities.”
Deconstructed Problem Blueprint
WHAT TO BUILD?
An edge-AI onboard processing software combined with a centralized cloud/web-based GIS dashboard that transforms public transport fleets into real-time urban sensing and monitoring units.
WHO IS IT FOR?
The problem belongs to Bharat Electronics Limited (BEL) and city municipal/transport authorities. The users include transport operations teams, road maintenance engineers, traffic police, and indirectly, citizens who benefit from safer and better-maintained roads.
WHY DOES IT EXIST?
Current urban management relies on sluggish manual inspections, fixed CCTV cameras with massive blind spots, and delayed citizen complaints. This results in slow maintenance responses, unmapped road defects, and reactive safety enforcement.
INPUT → PROCESS → OUTPUT ARCHITECTURE PIPELINE
Live multi-channel video streams from bus-mounted cameras (front, rear, side, cabin), GPS tracking metadata, timestamps, and vehicle telemetry data.
Edge inference using lightweight object detection and computer vision models for pothole detection, vehicle counting/classification, traffic sign missing-status identification, ALPR (Automatic License Plate Recognition), and spatial-temporal aggregation.
Actionable GIS-based maps, real-time traffic congestion heatmaps, infrastructure deficiency lists, incident alerts (rash driving, hit-and-run) with clipped evidence, and automated maintenance reports.
CONSTRAINTS
- Must run locally on low-cost edge hardware onboard the bus with high FPS.
- Intermittent or low-bandwidth cellular connectivity requiring smart offline data caching and compression.
- High variation in lighting conditions (night driving, harsh sunlight, heavy rain/fog).
- Real-time constraint for critical incident alerts.
JURY METRICS
- Edge-AI inference latency (Frames Per Second processed locally)
- Pothole and defect detection precision/recall accuracy
- ALPR extraction confidence score accuracy under motion blur
- Bandwidth usage optimization percentage via edge filtering
- Dashboard responsiveness and UI/UX clarity for authorities
CORE BOTTLENECK
Real-time processing of multiple simultaneous high-definition video streams on constrained edge hardware onboard a moving bus without dropping frames or overheating.
Why You Should Choose This PS (Pros)
- High social impact addressing critical urban governance and safety problems.
- Clear, highly modular deliverables dividing cleanly between edge-AI and central dashboard.
- Strong commercialization and defense-grade dual-use potential for BEL.
- Great visual appeal with live GIS maps and heatmaps during jury presentation.
Challenges & Risks to Watch (Cons)
- Heavy compute requirements for multi-camera video processing on the edge.
- Lack of standard public datasets explicitly tailored for bus-mounted multi-angle road defect and traffic tracking.
- Crowded domain if approached with generic object detection without domain-specific customization.