2020/05/05 by Salar Rahili, Rahili, Salar, Benjamin Rivière +3
Engineering · Social Sciences · #FOS: Mathematics #Optimization and Control (math.OC) #Traffic control and management #Transportation Planning and Optimization #Transportation and Mobility Innovations
paper · pdf · doi:10.48550/arxiv.2005.01976
openalex publication_date 2020/05/05 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
In this paper, a learning-based optimal transportation algorithm for\nautonomous taxis and ridesharing vehicles is presented. The goal is to design a\nmechanism to solve the routing problem for multiple autonomous vehicles and\nmultiple customers in order to maximize the transportation company's profit. As\na result, each vehicle selects the customer whose request maximizes the\ncompany's profit in the long run. To solve this problem, the system is modeled\nas a Markov Decision Process (MDP) using past customers data. By solving the\ndefined MDP, a centralized high-level planning recommendation is obtained,\nwhere this offline solution is used as an initial value for the real-time\nlearning. Then, a distributed SARSA reinforcement learning algorithm is\nproposed to capture the model errors and the environment changes, such as\nvariations in customer distributions in each area, traffic, and fares, thereby\nproviding optimal routing policies in real-time. Vehicles, or agents, use only\ntheir local information and interaction, such as current passenger requests and\nestimates of neighbors' tasks and their optimal actions, to obtain the optimal\npolicies in a distributed fashion. An optimal adaptive rate is introduced to\nmake the distributed SARSA algorithm capable of adapting to changes in the\nenvironment and tracking the time-varying optimal policies. Furthermore, a\ngame-theory-based task assignment algorithm is proposed, where each agent uses\nthe optimal policies and their values from distributed SARSA to select its\ncustomer from the set of local available requests in a distributed manner.\nFinally, the customers data provided by the city of Chicago is used to validate\nthe proposed algorithms.\n