A Bi-Objective Model for Agency Centers Location, Scheduling, and Routing in the Taxi Industry: Incorporating Split Simultaneous Pickup and Delivery to Maximize Passenger Satisfaction

This paper aims to enhance the efficiency and quality of service in the taxi industry by proposing a bi-objective mixed-integer programming (MIP) model. The model focuses on minimizing total costs while maximizing passenger satisfaction, incorporating agency centers' location, scheduling, and routing with split simultaneous pickup and delivery.
 
This study is developmental-survey in nature and employs a descriptive-survey approach for data collection. Methodologically, it falls under the Hard Operations Research classification. An exploratory descriptive approach has been applied, using a ladies' taxi agency affiliated with a university in Rafsanjan city as a case study. The proposed model was solved for a small-scale case study using the augmented ɛ-constraint method in CPLEX software, version 12.1. Additionally, medium- and large-scale instances were provided to evaluate the performance of the solution approach through the rolling horizon algorithm.
 
The proposed model was first tested on a small instance with 6 destination nodes and 2 hubs, and detailed solution results were presented. Three key factors were analyzed to evaluate their impact on real-world scenarios: satisfaction reduction point (SRP), waiting time to receive services (WT), and total time for each taxi (T). The Pareto frontier analysis for the bi-objective model across different WT values demonstrated that increasing WT improves responsiveness. Sensitivity analysis of the SRP parameter revealed that passengers with higher SRP experience lower dissatisfaction levels. Similarly, sensitivity analysis of the total time parameter (T) indicated that increasing T enhances network responsiveness while reducing transportation costs. To address larger instances, a mixed-integer programming (MIP) model and a rolling horizon heuristic were developed and tested on three groups of test problems. Results showed that the rolling horizon algorithm efficiently solved instances with 12 nodes to an optimality gap of less than 0.001% in about 3 minutes, compared to 15 minutes required by the MIP model. For larger instances with 20 and 30 nodes, the rolling horizon algorithm completed in 5 and 19 minutes, respectively, whereas the MIP model required significantly more time.
 
In today's rapidly evolving world of internet-based taxi services such as Snap and Tapsi, applying a bi-objective location, scheduling, and routing model with split simultaneous pickup and delivery can optimize both total transportation costs and passenger satisfaction. This paper employs an augmented -constraint method to solve the problem for small instances and conducts sensitivity analysis on key problem features. The resulting Pareto points provide decision-makers with a wider range of options. For large-scale instances, the rolling horizon algorithm is used to efficiently solve the problem. The main advantage of the proposed model lies in its applicability not only to taxi agencies but also to supply chains for perishable products and other service industries. While taxi agencies can directly benefit from the model’s results, future research could extend the approach to perishable products where spoilage begins simultaneously at time zero across all nodes.