Inspiration
Urban waste pickup is often inefficient and uneven. Some areas get serviced reliably while others are delayed or skipped. We wanted a robotics-style “decision brain” that reduces wasted travel while prioritizing high-need and underserved zones to support SDGs 11, 12, and 13.
What it does
Icarus: FairRoute computes routes for a fleet of waste/recycling robots under real constraints like capacity, battery/range, and time windows. It optimizes for low travel cost while penalizing missed pickups and improving fairness so coverage is more equitable across zones.
How we built it
We modeled the city as pickup nodes with demands, service times, time windows, and zone labels. Then we compared a greedy nearest-feasible baseline to an optimized solver-based planner. The optimized planner uses constraint-based routing with a multi-objective cost (distance + missed-pickup penalties + fairness penalties) and outputs routes plus metrics and visualizations.
Challenges we ran into
Our first “optimized” runs looked worse on total distance because the solver served far more stops and avoided penalties, making naive distance-only comparisons misleading. We fixed this by reporting the right metrics (served stops, missed stops, distance per stop, and zone fairness) and tuning penalties so results matched real-world priorities.
Accomplishments that we're proud of
We built a fully computational fleet planner that demonstrates robotics-grade decision-making without any sensors or hardware. We also added a fairness-aware objective and clear baseline-vs-optimized evidence with route plots and a metric table.
What we learned
Optimization is about trade-offs: a plan can travel farther yet be “better” because it completes more work and improves equity. Clear constraint modeling and the right evaluation metrics matter as much as the algorithm itself.
What's next for Icarus: FairRoute
We’ll replace synthetic geometry with real road-network distances, add explicit charging-stop logic, and calibrate energy consumption from real robot specs. We’ll also add demand forecasting so the system predicts which bins will overflow and routes the fleet proactively, not reactively.
Log in or sign up for Devpost to join the conversation.