Aston EcoFlow | Devpost

Inspiration

Our project was inspired by Birmingham’s Clean Air Zone and the city’s push toward sustainability. As locals, we see how transport decisions directly affect our community’s air quality and daily lives.

The hackathon theme of community made us question how transport infrastructure is planned. Cities often rely on outdated or averaged data, which ignores how people actually move and interact. We wanted to explore how greener transport could be designed from real community behaviour, not assumptions.

To do this, we built a digital simulation of Aston, where autonomous agents represent individuals in the community. Each agent makes its own decisions, but together they form a living picture of collective demand. By observing this shared behaviour, we can identify where infrastructure can better serve the community while reducing CO₂ emissions.

What it does

The system simulates a full day in Aston, with hundreds of agents following daily routines like commuting, errands, and returning home. As agents move, their travel intent is aggregated into a demand layer, showing where pressure builds across the city.

First, the model records a baseline where infrastructure does not adapt, measuring distance travelled, efficiency, and carbon emissions. Then, using the community’s own demand patterns, the system generates new transport corridors shaped by how people actually move. Running the simulation again with these routes shows how community-driven infrastructure can reduce unnecessary travel and significantly lower CO₂ emissions through efficient routes.

How we built it

There was a lot of stress at the start, but we managed to get solid brainstorming done and quickly built an early mock-up using Lovable to shape how the system could be structured. Having a visual backbone early on helped turn an abstract idea into something tangible, and from there the project really started to work.

We built the system as a full-stack application using Python, TypeScript, and React, combining real transport data from Transport for West Midlands with an agent-based simulation. On the backend, we process real bus stop locations and route geometry using GTFS data. On the frontend, React and Leaflet are used to visualise Aston, autonomous agents, demand flows, and show proposed transport corridors.

Each agent makes independent travel decisions which are linked to nearby real bus stops using a k-nearest-neighbour search allowing the movement spatially realistic. Individual travel intent is then aggregated into a demand graph, and a greedy optimisation algorithm generates new transport corridors that capture the most community demand with minimal infrastructure.

Challenges we ran into

Key challenges included integrating real-world data, designing believable agent behaviour at scale, and merging rapidly evolving frontend and backend systems under time pressure. Overcoming these challenges proved that this idea might be semi-finished by the end!

Accomplishments that we're proud of

We delivered a working prototype of a living digital city, where individual decisions combine into clear community-level insights. Turning abstract behaviour into measurable demand, efficiency, and CO₂ reduction was a major achievement.