Abstracts and Selected Posters/Papers from 2166 - AP010 - Evaluation and Optimization of Transit Operations and Management

Authors share research on optimizing transit management and passenger flow and improved asset management.

Full Details about the Poster Session can be found here: Evaluation and Optimization of Transit Operations and Management 

Optimized Passenger Redirection During Incidents in Urban Public Transportation Systems

Poster, Paper 

Authors: Frederik R. Bachmann, Florian Dandl, Arslan Ali Syed, Roman Engelhardt, Klaus Bogenberger, Technische Universitat Munchen

Incidents disrupt the public transportation (PT) operation daily causing parts of the PT system to be temporarily out of services. The operations control center (OCC) takes multiple dispositive measures to resolve the incident and to mitigate its negative effects on the PT services. Usually, such measures are supply-centric readjustments of PT lines. Recent studies have shown that passenger-centric extensions can further mitigate the negative effects of incidents. This paper presents a passenger-centric incident management method, in which the passengers directly affected by an incident are given a redirection path advice to systematically reduce the total delay of all the passengers. This advice is consistent for all passengers associated with the same origin-destination-relation. It is assumed that dispatchers in OCCs often have quite a good intuition on the duration of incidents and have access to demand estimations. Based on these assumptions this study compares a heuristic and an optimization-based approach. The procedure first simulates scenarios without incident with incident but only supply-centric measures (line-splitting, rerouting PT lines) to set the lower and higher benchmark for the overall delay, respectively. Then SUMO simulations evaluate the benefits of the additional path advice from the heuristic and optimization-based passenger-centric procedures. The results show that both approaches can significantly lower total overall passenger delay with optimization providing the best results.

Vehicle Type Selection for Single-Line Transit Fleet Replacement

Poster

Authors: Chunyan Tang & Ailing Zhao - Dalian Maritime University, Tao Liu - Southwest Jiaotong University, Jiyu Zhang - Chang'an University

Fleet replacement is an important decision-making activity in public transit planning and operations. Recently, with the goal of reducing carbon dioxide emissions and achieving carbon neutrality, diesel buses are being replaced with new electric buses (EBs) in many cities around the world. The study proposes a methodology to analyze and compare the life-cycle cost of multi-type EBs, including purchase cost, operating cost, charging infrastructure cost, as well as passenger waiting time. The proposed methodology is applied to a real-life case study of a bus line in Dandong, China. Three different types of EBs, i.e., small-, middle- and large-size EBs, with fast and slow charging technologies, are considered in the analysis. The results show that no matter which charging technology is considered, the life-cycle cost of large-size EBs is the lowest. From the perspective of passengers, small-size EBs that result in shorter waiting time are more favorable. When the increase rate of the total passenger demand is the same as that of the max-load passenger demand, the total passenger waiting time for each vehicle type will keep fixed. The results also indicate that the bus line length does not have a significant impact on EBs vehicle selection. In addition, vehicle purchase subsidies will not affect EBs vehicle selection due to introducing a subsidy reduction policy.

Optimal Investment Timing and Length Choice for Rail Transit Under Stochastic Demand

Poster

Authors: Qianwen Guo - Florida International University, Shumin Chen - Guangdong University of Technology

This paper extends the public transit investment literature by jointly optimizing the investment timing and rail length choice under future demand uncertainty. First, a deterministic dynamic model is proposed to find the optimal investment timing and optimal rail length with the objective of minimizing the total system cost. Moreover, we assume that the source of uncertainty comes from the demand density, which follows a geomatric Brownian motion with jumps of random magnitude, occurring in ramdom times, according to a Poisson process. Analytical solution for the optimal investment timing is presented under the stochastic demand. The impact of the demand shock along with other uncertainties on the optimal demand thresholds have been analyzed. Several distributions for these jumps are considered and compared regarding the optimal investment timings. Numerical results show that: (a) demand uncertainty leads to an investment delay and an increase in the optimal rail length; (b) Determinisitc and constant negative jumps and the same magnitude of the jump distributed according to an exponential distribuction will leads to early investment compared with the case without jumps. (c) The investment timing and rail length choice under the stochastic case with unexpected conjuncture shock is larger than stochastic case without jumps and larger than the deterministic case.

Machine Learning Predictive Model for Small Urban and Rural Transit Systems to Achieve and Maintain Public Transportation Rolling Stock in a State of Good Repair

Authors: Dilip Mistry - Transportation Safety Systems Center, Jill Hough - North Dakota State University

Achieving and maintaining public transportation rolling stock in a state of good repair (SGR) is crucial to providing safe and reliable services to riders. Transit agencies that utilize federal grants are required to maintain their transit assets in a state of good repair or operating at a full level of performance. Therefore, transit agencies in small urban and rural transit systems would benefit from an intelligent predictive model for analyzing their transportation rolling stock, determining the current conditions, predicting when they need to be replaced or rehabilitated, and determining the funding needed to replace them at some point in the future to maintain the state of good repair. Many transit agencies in small urban and rural transit systems do not have analytical tools for predicting the service life of vehicles, the backlog of vehicles beyond their service lives, and the replacement costs of vehicles. Consequently, this simple predictive model along with a financial analytical tool developed from the output of the predictive model would be a valuable resource for their state of good repair needs and their prioritization of capital needs for replacement and rehabilitation.

A Reinforcement Learning Framework for Passenger Flow Control Based on Asynchronous Advantage Actor-Critic Recurrent Neural Network

Authors: DBao Wang & Qiming Su - Southwest Jiaotong University School of Transportation and Logistics, Luo Xia - Southwest Jiaotong University, Peter Jin - Rutgers University

Passenger flow control is an effective way to alleviate congestion, improve safety and service quality in metro network. However, it would be challenging to build and solve the control model considering the dynamic travel demand and complex metro network structure, and it’s also difficult to put into practice due to the substantial computation costs. Under this circumstance, this study proposed a general framework for passenger flow control based on reinforcement learning (RL), which includes the network state characterization unit, the control model unit, and the reinforcement learning unit. Firstly, referring to the definition in reinforcement learning, the “action”, “state” and “reward” are respectively described as decision variables, constraints and objective functions in the built passenger flow control programming model. Secondly, an interacting mechanism is elaborated to coordinate the control schemes provided by the reinforcement learning unit and the network states, which is evolved in the network state characterization unit iteratively. Thirdly, the Asynchronous Advantage Actor-Critic Recurrent Neural Network (A3C-RNN) is trained to solve the tricky programming model as it efficiently utilizes computational resources and temporal information. Finally, the proposed framework is validated by a case study using the data from Chengdu Urban rail transit (URT). The results show that the proposed method is feasible when considering different objectives such as minimizing passenger waiting time, maximizing passenger turnover and maximizing passenger number.