OPTIMIZATION OF MEDAN CITY WASTE TRANSPORTATION SYSTEM USING MULTIPLE-TRIP VEHICLE ROUTING PROBLEM (MTVRP) MODEL AND SIMULATED ANNEALING

Medan generates approximately 2,000 tons of waste daily, yet only 800 tons are successfully transported to landfills, indicating significant inefficiencies in waste transportation. This study addresses the issue by applying the Vehicle Routing Problem with Multiple Trips (VRPMT) combined with the Simulated Annealing (SA) algorithm to optimize waste transport operations. The VRPMT model allows each vehicle to make multiple daily trips, enhancing fleet utilization while ensuring that all service points are visited, vehicle capacities are not exceeded, and vehicles return to the depot after each trip. The study focuses on Tegal Sari Mandala II (TSM II), Medan Denai, a densely populated neighborhood with narrow roads that require bestari pedicabs for flexible waste collection. Data includes waste collection points, vehicle capacities, transport frequencies, and operational costs. The SA algorithm begins with a random route solution, then iteratively evaluates and improves it by minimizing total distance and cost. It also avoids local optima through a controlled temperature reduction process. Results demonstrate significant improvements: total travel distance was reduced from 12,500 meters to 8,646 meters (a 30.8% reduction), and operational costs decreased from IDR 12,000 to IDR 8,946 (a 25.5% reduction). On average, each bestari pedicab completed two daily trips, maximizing capacity utilization and minimizing penalty costs. The system integrates a structured database and Google Maps API for route visualization, enhancing planning and monitoring. Overall, this approach contributes to more efficient, cost-effective, and environmentally friendly waste transportation. It supports climate action goals and provides a scalable, replicable model for sustainable urban waste management in other regions facing similar logistical challenges. However, this study has some limitations. The VRPMT model was applied only in a neighborhood with a limited vehicle type, which may reduce its generalizability to broader urban areas with more complex logistics. Also, the Simulated Annealing algorithm settings were manually tuned and not benchmarked against other metaheuristic methods. Future studies could improve the model by considering dynamic traffic conditions, integrating real-time data, or testing hybrid optimization approaches to enhance its effectiveness and adaptability.