How it works
Two online rounds, then the finale at the Transport Stack annual event on 11 December 2026.
| Registration | Until 20 Nov | Register your team on Where U Elevate by 20 November 2026, 23:59 IST. |
|---|---|---|
| Phase 1 | Two weeks | Open to all registered teams. Pick Problem 1 or 2 and submit a proposal and a working prototype. |
| Phase 2 | Two weeks | For shortlisted teams. Refine your Phase 1 solution, or take on Problem 3, 4 or 5. Each team meets mentors for 10–15 minutes every week. |
| Finale | 11 Dec | Phase 2 solutions are presented and judged at the Transport Stack annual event, and winners are announced. |
Problems
Phase 1, open to all teams
Phase 2, shortlisted teams
Problem 1
Real-time bus traffic model and journey planner
Background. Delhi's buses generate continuous location data across the city's road network. Combined with historical patterns, this data can model traffic flow and travel times across the network, making the bus fleet a strong proxy for services like Google Maps.
Challenge. Build a traffic model from real-time and historical bus location data and integrate it with a journey planning interface. The goal is a Google Maps-like API that estimates bus travel times between any two locations.
Data. Historical and real-time bus location data for Delhi: GPS traces, route information and stop locations.
Success criteria. A working API that takes two locations and returns plausible bus journey options with estimated travel times. Estimates should hold up against known actual travel times for sample location pairs and against Google Maps.
Problem 2
Crowdedness-aware journey planner
Background. Crowdedness shapes how commuters experience a trip, yet existing journey planners ignore it. E-ticketing and Electronic Ticketing Machine (ETM) data make it possible to estimate how crowded buses are across routes and times of day.
Challenge. Build a model that estimates crowdedness across bus routes and times of day from about 4–5 weeks of e-ticketing and ETM data, then integrate it into a journey planner so commuters can choose routes on both travel time and expected crowdedness.
Data. E-ticketing and ETM data for 4–5 weeks, covering ticket sales, routes and timestamps. Route and stop information from Transport Stack.
Success criteria. A journey planner that shows crowdedness estimates alongside route options. Estimates should be grounded in the ticketing data and hold up when spot-checked against known peak and off-peak patterns.
Problem 3
Origin–destination matrix for Delhi public transit
Background. Knowing where trips start and end is the foundation of transit planning, from route design to service frequency. Delhi has two partial views of this: bus e-ticketing data, which captures only 3–4% of bus tickets, and metro gate data, which is far more complete.
Challenge. Estimate an origin–destination (OD) matrix for Delhi's public transit network that matches real-world travel as closely as possible. The core problem is extrapolation: scaling a small bus e-ticketing sample to the full ridership, using the denser metro data to inform travel density across the city.
Data. Bus e-ticketing data, metro entry/exit gate data, and route and stop information from Transport Stack.
Success criteria. A reproducible OD matrix at stop or zone level, with a documented extrapolation method. Results should be consistent with known ridership totals and with metro gate counts, and the method should state its assumptions and where uncertainty is highest.
Problem 4
Feeder bus routes for metro stations
Background. Many metro riders struggle with the first and last mile between the station and their home or workplace. A well-planned feeder network can extend the reach of every station, but only if routes fit real road conditions and the operating limits of an electric fleet.
Challenge. Design feeder routes for a given list of metro stations using a fleet of about 500 electric buses, maximising coverage. Use metro entry/exit data to decide where demand is, and plan around charging needs. Routes must meet these constraints:
- Maximum route length of 7 km (local service only)
- Account for traffic conditions
- Use only roads wider than 6 metres on each side
Data. List of target metro stations, metro entry/exit data, road network data, and bus traffic and stop information from Transport Stack.
Success criteria. A set of feeder routes with fleet allocation that respects every constraint, along with clear coverage metrics such as the share of population or metro demand served. The charging plan should be feasible for the proposed schedules.
Problem 5
Metro and bus crowding estimation
Background. Automatic Fare Collection (AFC) systems record where passengers board and alight. Across a full month, this data can show how passenger load builds and falls along each line and route, which operators can use to plan capacity and commuters can use to plan trips.
Challenge. Estimate crowding levels on Delhi's metro and buses from historical AFC boarding and alighting data. Produce load estimates by line or route, segment, and time of day.
Data. About one month of AFC boarding and alighting data, along with route, stop and station information from Transport Stack.
Success criteria. Crowding estimates by segment and time band that can be queried or visualised. The method for turning boarding and alighting records into on-vehicle load should be documented, and results should match known peak and off-peak patterns.