SmartPark OGF
A predictive parking proposal for One Galle Face that combines demand forecasting, fair prorated billing, computer-vision occupancy insights, and an offline-capable driver experience.
The Challenge & Core Problem
High-Level Briefing
The proposal describes One Galle Face visitors facing high parking charges, limited guidance, no advance occupancy prediction, and billing that rounds partial hours upward. It cites a documented Rs. 750 charge for a 2-hour-15-minute visit.
"Design a predictive parking ecosystem that forecasts demand 30 minutes ahead, allocates capacity across three parking levels, applies fair prorated billing, and supports an offline-capable driver experience."
Hard Timebox & Rules
- The zero-budget proposal uses simulated IoT data rather than installing physical parking sensors.
- The system combines a predictive engine, agent-based decision flow, computer vision, and operational dashboards.
- The driver-facing experience is designed as an offline-capable Progressive Web App.
Open Risks & Pressure Factors
- Performance figures in the proposal are projected targets, not field-measured results.
- The proposal relies on simulated data for evaluation and does not document a live OGF deployment.
Leadership & Personal Contribution
Team Leadership
Real-Time Architectural Dilemmas Under Clock Pressure
The Technical Solution
SmartPark OGF combines 30-minute demand forecasts, a six-agent decision loop, a parallel computer-vision agent, operational dashboards, and fair prorated billing in an offline-capable parking concept.
Data Flow & Architecture
The proposal sends simulated IoT and CCTV-derived telemetry through a stream processor and predictive engine. Monitor, Planner, Validator, Action, Dashboard, and Feedback functions form the decision loop, while a Vision Agent supplies occupancy, queue, and anomaly signals. Outputs include a driver PWA, mall dashboard, signage, and billing flow.
Critical Data Pipelines
The Retrospective: What Worked & What Needs Work
What Worked
- ✓The proposal ties the concept to documented parking complaints and a specific receipt example.
- ✓The architecture connects forecasting, operational decisions, computer vision, and visitor-facing guidance.
- ✓The design includes offline support and prorated billing as explicit product requirements.
What Still Needs Work
- ✕The proposal presents target performance figures as projections rather than measured results.
- ✕The documented work does not establish a live deployment or validation with physical parking sensors at OGF.
Hard-Won Engineering Takeaways
Predictive Parking and the Agent Decision Loop
The proposal combines short-horizon demand forecasting with agent-based policy checks and a parallel vision layer, then routes approved actions to both mall operations and drivers.