A Hybrid Approach Based on Machine Learning and Optimization Techniques for Waste Management Collection Using Electric Vehicles

Simple item page
AuthorHejres, Shaima
metadata.dc.contributor.departmentUniversity of Bahrain, College of Information Technology, Computer Science
metadata.dc.date.accessioned2026-09-20T15:54:08Z
Publication date2026
Degree levelDoctoral
AbstractAbstract : In this era, the growing demand for Electric Vehicles (EVs) depends on the high inte-gration of this technology in many areas. Thus, an important area of research raises interest in finding the optimal path-planning solution for EVs. However, the Waste Management System (WMS) is an important economic sector that helps to support health and improve the quality of life in smart cities. Hence, this thesis discusses research for EV; specifically, Unmanned Aerial Vehicle (UAV), routing techniques to assist researchers in choosing the appropriate approach for developing a system based on optimization techniques, machine learning, or hybrid approaches in managing a waste collection system (WCS). Hybrid ap-proach techniques that combine optimization/genetic algorithm (GA) and machine learn-ing/reinforcement learning (RL) techniques can contribute to the production of effective solutions to problems for WCS, which is a vital part of WMS, by increasing the robust-ness in solving UAV routing problems. In this thesis, we address the problem of waste collection using UAVs, with the primary objective of minimizing both the collection time and energy consumption. Three different approaches are proposed. The first two focus on a single UAV, while the third involves the coordination of multiple drones. The first ap-proach employs a multi-objective genetic algorithm. The second approach integrates GA with RL in a hybrid model. The third approach combines GA, RL, k-means clustering, and the A* algorithm to coordinate multiple UAVs. Across all approaches, several opera-tional constraints are taken into account, including UAV capacity, flight time, and others. The multiobjective GA approach achieved a significant error reduction compared to previ-ous research using the same experimental parameters: more than 81% for 100 instances, more than 59% for 500 instances, and over 44% for 1,000 instances. The second approach, which applies RL following genetic optimization, further improved performance by an av-erage of 14.35% compared to the GA only approach. For the multi-UAV strategy (third approach), two scenarios were developed: one with UAV reassignment and one without. The results showed that the UAV reassignment significantly improved performance, reduc-ing flight time by 32.11%, energy consumption by 63.12%, and the total distance traveled by 31.14% compared to the scenario without reassignment.
metadata.dc.identifier.otherDrupal node 4618
metadata.dc.identifier.otherhttps://digitalrepository.uob.edu.bh/en/dar/hybrid-approach-based-machine-learning-and-optimization-techniques-waste-management-collection
metadata.dc.identifier.urihttp://165.227.56.122/handle/123456789/7544
Languageen
Linkhttps://digitalrepository.uob.edu.bh/en/_flysystem/fedora/2026-02/002479_a_hybrid_approach_based_on_machine_learning_and_optimization_techniquesfor_waste_management_coll.pdf
TitleA Hybrid Approach Based on Machine Learning and Optimization Techniques for Waste Management Collection Using Electric Vehicles
TypeDoctoral Dissertation

Files

Original bundle

Now showing 1 - 1 of 1
Thumbnail Image
Name:
node-4618-1-002479_a_hybrid_approach_based_on_machine_learning_and_optimization_techniquesfor_waste_management_coll.pdf
Size:
5.84 MB
Format:
Adobe Portable Document Format
Download

Collections

Doctoral Dissertations