A Multi-Objective Ensemble Model to Suggest an Optimal Path for Private Ride-Hailing Services

Optimal path suggestion for taxi cabs being operated by private Fleet Management Services (FMS) seems a classical combinatorial problem, yet attracts researchers when more conflicting objectives are met. To improve the Quality of Service (QoS), the FMS employed a variety of strategies, compromising on some conflicting objectives, such as reducing passenger waiting time by having cabs cruise on roads, thereby increasing traffic, air, and noise pollution. To find an optimal solution for such kind of objectives, herewith a multi-objective optimization model is proposed. This model fuses the best features of Bio-inspired algorithms with Recurrent Neural Networks (RNNs) and Graph Neural Networks (GNNs), providing optimal paths for FMS cabs. The proposed model has been trained and tested on the 2024 Chicago Taxi trips and Traffic datasets, which were downloaded from the official Chicago government portal.s The Points of Interests (POIs) were crawled using the Google Maps API. Results showed that the proposed model outperformed the baseline model by reducing CO2 emissions by an average of 11.4%. In addition, the time taken to pick up is also reduced to a total of 3.70 km for the spatiotemporal parameters taken. These reductions indicate that the proposed model has minimized unnecessary cruising time, waiting time, and CO2 emissions to a greater extent. This model is modular, scalable, and provides balanced revenue for the FMS that runs in a resource-constrained environment.