Optimizing the bus operation plan Based on Deep Learning

Abstract

The bus operation plan is closely to urban resident daily life. A good bus operation plan can improve the travel efficiency and comfort at the same time. To reduce the cost of traveling with bus companies and passengers, here a bus operation model considering shuttle buses based on predicted data (PSBO) is proposed. This model contains three parts, namely the passenger flow prediction, bus travel time prediction, and bus operation optimization. In the passenger flow prediction part, we propose a novel passenger flow prediction model based on long short-term memory (LSTM) and the passenger travel frequency, called Two-stage LSTM passenger flow prediction model (TLPLP). The accuracy of the TLPLP model is verified by comparison with multiple models. This paper also chooses the LSTM model to predict the bus travel time between different stations. The predicted passenger number and bus travel time are two inputs in the bus operation optimization part of PSBO model. This model aims at minimizing the passenger travel time and the operation cost through an insert shuttle buses method. We design a quick algorithm to solve this model and verifies the effect of this model compared with the bus dispatching plan in use.

Introduction

Public transport is a system of transport for passengers by group travel systems, typically managed on a schedule, operated on established routes, and that charge a posted fee for each trip. As a major part of public transport, the bus operation systemplays an important role. Ways to satisfy passenger demand and reduce the travel costs are always the research focuses of scholars. The bus operation plan is usually divided into two parts: traffic demand prediction and bus operation optimization.

In the authors' daily life, we found that there are often few passengers in some buses for line 87 in the morning peak. The peak bus organization for this line is a combination of regular buses and shuttle buses. So, this paper proposed a bus operation plan model considering the shuttle bus based on predicted data (PSBO) to solve this problem. This model contains three parts: the passenger flow prediction, the bus travel time prediction, and the bus operation optimization.

In the passenger flow prediction part, we proposed a novel passenger flow prediction model based on long short-term memory (LSTM) and the passenger travel frequency, called Two-stage LSTM passenger flow prediction model (TLPLP). In the bus travel time prediction part, we used the LSTM model to obtain bus travel time between different stations. In the bus operation optimization part, this paper built an optimization model to minimize the passenger time cost on the condition of meeting traffic demand by inserting the shuttle buses. We designed a quick algorithm to solve this problem. We drew a bus working diagram based on the solution and made a comparison between different bus capacities. The structure of this paper is as follows. Section 1 introduces the research purpose of the paper. Section 2 describes the latest research methods in this area and our differences with them. Section 3 describes the PSBO model framework and its sub-models. Results of the experiments are provided in Section 4. Conclusions are shown in Section 5.