It's time to optimize how robots serve cities

A sophisticated Vehicle Routing Problem with Time Windows (VRPTW) solver using Google OR-Tools. The system optimizes routes for autonomous waste collection robots while enforcing capacity limits, time windows, and battery constraints. Unlike simple greedy algorithms that pick the nearest stop, our solver uses Guided Local Search—a metaheuristic that escapes local optima through penalty-based diversification. The fairness-aware penalty system ensures underserved neighborhoods aren't neglected, while global span cost coefficients balance workload across the robot fleet.

Traditional routing optimizes only for distance, neglecting fairness across neighborhoods. Icarus uses penalty-based disjunctions to prioritize underserved zones, ensuring no community is left behind. The system measures fairness using zone completion rates, disparity metrics, and Gini coefficients for waiting times.

The solver handles multiple simultaneous constraints: capacity limits (kg per robot), time windows (service hour restrictions), battery/range limits (max distance per charge), and shift duration constraints. It uses global span cost coefficients to balance workload across the fleet, preventing one robot from being overworked while others idle.

Vehicle routing with constraints is a canonical AI problem (NP-hard combinatorial optimization). The system uses Guided Local Search as its primary metaheuristic, which iteratively improves solutions through penalty-based diversification. It also supports Simulated Annealing and Tabu Search for comparative analysis.