Inspiration
Waste collection follows a schedule, but waste generation doesn’t. Across a campus, some bins fill quickly while others remain nearly empty. Collecting both on the same fixed route can mean unnecessary driving while high-demand locations still need attention.
We built Flux around a simple question: What if waste infrastructure could respond to its actual condition? We wanted to connect sustainability with a practical operational decision: helping operators see which bins need service and plan collection accordingly.
What it does
Flux is an interactive digital twin of a campus waste network. It brings bins, trucks, collection stations, and simulated fill-level telemetry together on one map.
Operators can:
- Explore historical days and filter trash, recycling, and compost bins.
- Identify bins at or above the 80% pickup threshold.
- Select bins for collection and generate multi-truck pickup routes.
- View road-following routes, estimated travel times, and collection stops.
- Compare planned collection with a daily baseline using estimated mileage, fuel consumption, and CO₂ emissions.
Our environmental metrics are estimates based on route distances and vehicle assumptions, rather than measured exhaust emissions.
How we built it
We built Flux with HTML, CSS, and JavaScript, using Leaflet and OpenStreetMap for the interactive campus map. CSV uploads provide bin locations, fill readings, truck states, collection stations, and optional event information.
We also used OpenAI and Claude as development assistants to help with coding, debugging, and refining implementation ideas.
Our routing logic uses a greedy assignment approach: it repeatedly assigns a remaining pickup to the truck whose current position is closest, then updates that truck’s position. We use OSRM to turn the resulting pickup sequences into road-following routes with distance and travel-time estimates.
The application stores saved routes separately for each day and compares their combined mileage and estimated fuel use against one daily baseline. Fuel estimates use truck MPG, and CO₂ calculations extend that comparison into an environmental metric.
We designed the interface around a calm visual identity, clear fill-level colors, and accessible route controls.
Challenges we ran into
One challenge was connecting geographic proximity with realistic driving routes. Two locations can look close on a map while requiring a longer trip along roads. We addressed this by separating pickup assignment from road routing and including a straight-line fallback when the routing service is unavailable.
Another challenge was keeping comparisons consistent across multiple saved routes. Counting the baseline again for every route would exaggerate savings, so we store one baseline per day and compare it with the combined collection effort.
We also needed to bring several datasets into one coherent interface while keeping simulated readings and estimated environmental metrics clearly distinguishable from real-world measurements.
Accomplishments that we're proud of
We’re proud of building a connected workflow from waste telemetry to a usable collection plan. Operators can inspect conditions, select pickups, generate multi-truck routes, and review estimated operational and environmental impact within the same interface.
We’re also proud of making the system understandable visually. Fill-level markers, waste-stream filters, road-following routes, and saved route summaries turn separate datasets into information an operator can act on.
Adding CO₂ estimates helped connect our technical work to the sustainability goal that inspired Flux.
What we learned
We learned that infrastructure software needs to connect data with decisions. A fill reading becomes useful when it helps someone determine where to send a truck and understand the consequences.
We also learned that routing involves more than finding nearby points. Road geometry, truck locations, collection stations, and multiple trips all affect the resulting plan.
Finally, we learned how much assumptions matter when communicating impact. Mileage, fuel, and emissions estimates are useful, but their value depends on explaining how they were calculated and what the comparison represents.
What's next for Flux
Our next step is building low-cost bin sensors using ultrasonic measurements and microcontrollers, then connecting them to a backend so the dashboard can receive live readings instead of uploaded CSVs.
We also want to incorporate campus events directly into planning. Football games, concerts, and other gatherings could help explain changes in waste generation and eventually support forecasts of where additional service will be needed.
From there, we want Flux to predict future overflow and compare possible responses, including earlier pickups, combined collections, temporary capacity, or waiting. Operators would review the recommended response before committing resources.
We also plan to add truck-capacity and waste-stream constraints, strengthen forecasting with more representative data, and validate routing performance through equal-service comparisons. A campus pilot would help us measure whether Flux can reduce unnecessary collection effort while maintaining reliable service.
Log in or sign up for Devpost to join the conversation.