@inproceedings {pub6145,
	title = {Search Space Reduction Through Machine Learning for the Electric Autonomous Dial-A-Ride Problem},
	author = {Maria Bresich AND Guenther Raidl AND Caspian Coleman AND Pascal Welke AND Steffen Limmer},
	year = {2026},
	month = {January},
	abstract = {The complexity of combinatorial optimization problems often leads to a steep performance decrease of exact as well as heuristic solving approaches with increasing problem size. This paper explores the usage of machine learning to reduce the practical complexity of such problem instances by predicting and removing unpromising parts of the search space in a preprocessing step in order to accelerate the subsequent solving process. This approach is investigated for the electric autonomous dial-a-ride problem (E-ADARP), where self-driving vehicles that are powered by electricity are used to provide an efficient and sustainable ride-sharing service. The electric autonomous vehicles (EAVs) serve customer transportation requests between pickup and drop-off locations within specified time windows and potentially accommodate multiple customers at once. The more requests and vehicles are considered, the more increases the computational cost for planning minimum cost routes for the EAVs. We present features distinguishing high-quality areas of the search space and two variants of a support vector model which are tested regarding their individual predictive power as well as evaluated in terms of the resulting speed-up of a state-of-the-art algorithm for the E-ADARP. Results show that the proposed approach achieves substantial run time reductions of the solving process while barely reducing the solution quality or even
improving it. },
	publisher = {Springer, LNCS, LNAI, LNBI},
	booktitle = {International Conference on Machine Learning, Optimization, and Data Science (LOD) 2025}
}