Vatsal Tripathi
Department of Management, Birla Institute of Technology and Science, Pilani, Pilani Campus, Rajasthan, India.
Saurabh Chadha
Department of Management, Birla Institute of Technology and Science, Pilani, Pilani Campus, Rajasthan, India.
Satyendra Kumar Sharma
Department of Management, Birla Institute of Technology and Science, Pilani, Pilani Campus, Rajasthan, India.
DOI https://doi.org/10.33889/IJMEMS.2026.11.5.082
Abstract
Collection of MSW generally accounts for 60-80% of total budgets required for waste management services. Just take into account the fact that the urban centers in India alone generate more than 170,000 tons of MSW on a daily basis (CPCB, 2022). Thus, any inefficiency of collection routes translates into money losses. The Capacitated Vehicle Routing Problem (CVRP) offers a proven method for optimizing these routes. However, because it's an NP-hard problem, finding exact solutions for large urban areas is simply too computationally intensive. That's why metaheuristic algorithms have become the go-to solution for real-world waste collection challenges. Although there have been many applications of metaheuristic algorithms to CVRPs, there is a paucity of studies comparing the performance of various algorithms. This is because many researchers compare a few numbers of algorithms or even just two algorithms, while some perform single-run tests, which do not take into consideration the inherent randomness in searching using metaheuristics. It can also be noted that very few studies focus on examining how algorithms perform as the problem size increases. To bridge these gaps, our study conducted a controlled experiment. We ran 120 executions, testing four algorithms against three CVRPLib Augerat set a benchmark instances (n = 32, 48, and 80). Each configuration had ten independent runs. For every run, we simultaneously measured solution quality, computational speed, run-to-run consistency, and stagnation behaviour. The results presented clearly indicate that there is no single algorithm that is able to outperform others in all aspects of evaluations. ACO emerged as the most stable algorithm with a coefficient of variation below 2%. Thus, ACO can be used in routine and daily operations because of its stability. SA was shown to be the fastest one among the algorithms; it finished optimization within 1.4 seconds even on the largest sample considered. However, it is important to note that the geometric cooling schedule of SA (α=0.95) resulted in the loss of probabilistic search capabilities after about 135th iteration. GA and PSO showed high levels of variability. Overall, our results point to a scale-dependent shift in algorithmic performance around the medium-sized benchmark instance (A-n48-k7). Below this threshold, SA offers superior speed. Beyond it, ACO takes the lead in both solution quality and consistency. The main contribution of this research is in the form of Heuristic Selection Framework based on empirical data, which recommends SA for areas having less than 40 collecting bins and ACO for zones that have more than 60. This research has immense practical importance in terms of municipal corporations, procurement in smart cities, and logistics companies involved in waste management.
Keywords- CVRP, Metaheuristic comparison, ACO, Simulated annealing, Urban waste routing, Algorithm robustness, Smart city logistics.
Citation
Tripathi, V., Chadha, S., & Sharma, S. K (2026). A Cross-Architectural Comparison of Metaheuristic Algorithms for Capacitated Vehicle Routing in Urban Waste Collection Systems. International Journal of Mathematical, Engineering and Management Sciences, 11(5), 2060-2078. https://doi.org/10.33889/IJMEMS.2026.11.5.082.
