OPTIMASI PENGIRIMAN LAST MILE OLEH KURIR MITRA MENGGUNAKAN K-MEANS DAN ANT COLONY OPTIMIZATION

Abstract

Pertumbuhan pesat e-commerce setelah pandemi meningkatkan kebutuhan layanan logistik yang cepat, efisien, dan berbiaya rendah. Salah satu tahapan paling rumit dan mahal adalah last-mile delivery, yaitu pengiriman paket dari pusat distribusi akhir ke pelanggan. Kompleksitasnya dipengaruhi oleh banyaknya tujuan pengiriman, variasi jarak tempuh, serta keterbatasan jumlah kurir dan kendaraan. Untuk meningkatkan efisiensi operasional, perusahaan logistik menerapkan model kemitraan dengan kurir independen yang dibayar berdasarkan jumlah paket yang berhasil dikirimkan. Namun, model ini berpotensi menimbulkan ketimpangan pendapatan apabila pembagian paket dan area kerja tidak dilakukan secara adil. Pada salah satu perusahaan logistik di Cilebut, Kabupaten Bogor, terdapat 41 kurir mitra dengan rata-rata 822 paket per kurir, tetapi jumlah paket berkisar antara 96 hingga 1.215 paket, yang menunjukkan ketimpangan beban kerja yang tinggi. Penelitian ini membahas Task Allocation Problem, yaitu pendistribusian tugas secara efisien sesuai sumber daya yang tersedia. Dalam last-mile delivery, tugas tersebut berupa pembagian area dan paket kepada setiap kurir. Berbagai penelitian umumnya menggunakan strategi cluster first and route second melalui metode klasterisasi berbasis jarak, seperti K-Means dan Clustering-Based Routing Heuristic (CRH), yang dipadukan dengan optimasi rute menggunakan Vehicle Routing Problem with Time Windows (VRPTW) dan Ant Colony Optimization (ACO). Penelitian ini memperkenalkan pendekatan fairness-aware spatial task allocation yang menjadikan pemerataan beban kerja kurir sebagai hard constraint. Kontribusi utama penelitian terletak pada formulasi optimasi yang menyeimbangkan efisiensi melalui minimasi total jarak tempuh dan fairness melalui minimasi ketimpangan jumlah paket antar kurir. Konflik antara kedua tujuan tersebut diatasi menggunakan Capacity Constrained Voronoi Diagram (CCVD), yang menjamin seluruh paket teralokasi, setiap paket hanya ditangani satu kurir, dan setiap area pengiriman tetap terkoneksi. Setelah area seimbang terbentuk, ACO digunakan untuk mengoptimalkan rute pengiriman pada setiap area. Tahapan penelitian meliputi pengumpulan dan pra-pemrosesan data, pengembangan model optimasi area dan rute menggunakan K-Means, CCVD, dan ACO, serta implementasi dan evaluasi hasil optimasi. Evaluasi dilakukan dengan membandingkan jumlah paket dan total jarak tempuh setiap kurir sebelum dan sesudah optimasi. Penelitian ini diharapkan menjadi rekomendasi bagi perusahaan logistik dalam membangun model last-mile delivery berbasis kemitraan yang mendukung pemerataan beban kerja kurir mitra sekaligus menjaga efisiensi operasional perusahaan.

The rapid growth of e-commerce after the COVID-19 pandemic has increased the demand for fast, efficient, and cost-effective logistics services. One of the most complex and expensive stages is last-mile delivery, which involves delivering parcels from the final distribution center to customers. This stage is challenging because of the large number of delivery destinations, varying travel distances, and limited courier and vehicle resources. To improve operational efficiency, many logistics companies employ independent partner couriers who are paid based on the number of parcels successfully delivered. However, this payment model may create income inequality when delivery areas and workloads are not distributed fairly. Data from a logistics company in Cilebut, Bogor Regency, involving 41 partner couriers, revealed a significant workload imbalance, with an average of 822 parcels per courier and workloads ranging from 96 to 1,215 parcels. This study addresses the Task Allocation Problem, which focuses on efficiently assigning delivery areas and parcels according to available resources. Existing last-mile delivery approaches commonly adopt a cluster-first, routesecond strategy using distance-based clustering methods such as K-Means and the Clustering-Based Routing Heuristic (CRH), followed by route optimization using the Vehicle Routing Problem with Time Windows (VRPTW) and Ant Colony Optimization (ACO). This research proposes a fairness-aware spatial task allocation model that treats workload balancing as a hard constraint rather than a secondary objective. Instead of emphasizing a new algorithm combination, the main contribution is an optimization formulation that simultaneously minimizes total travel distance and workload disparity among couriers. Because K-Means can reduce travel distance while increasing workload imbalance, the proposed approach first applies the Capacity Constrained Voronoi Diagram (CCVD) to ensure that all parcels are assigned, each parcel is handled by only one courier, and every delivery area remains spatially connected. Ant Colony Optimization (ACO) is then used to optimize delivery routes within each balanced area. The research methodology includes data collection, data preprocessing, optimization of delivery areas and routes using K-Means, CCVD, and ACO, followed by implementation and evaluation. Performance is evaluated by comparing parcel distribution and the total travel distance of each courier before and after optimization. The proposed model is expected to provide practical recommendations for logistics companies in developing partnership-based last-mile delivery systems that promote fair workload distribution while maintaining operational efficiency.