Urban Rail Service Design for Collaborative Passenger and Freight Transport
Zhujun Li, Beijing Jiaotong UniversityShow Abstract View Presentation
Amer Shalaby, University of Toronto
Matthew Roorda, University of Toronto
Baohua Mao, Beijing Jiaotong University
This paper develops an operational strategy in which urban rail transit is used for freight transport. An environment-friendly urban freight transportation alternative using passenger rails is analyzed by employing optimization techniques to support the collaborative transportation of passengers and freight. Practical cases are investigated to test the technical feasibility of this transportation scheme. The paper formulates the train service design problem on a single urban rail line with passenger and freight. Passenger trains have a prescribed service frequency and timespan. Freight can be transported by inserting dedicated freight trains or utilizing the extra space inside the passenger train carriages. Station platforms are allowed to load and unload goods and passengers at the same time. An optimization model for combined train service design is proposed to maximize profit resulting from the balance of revenues and costs brought by the freight service. Efficient schedules of freight trains and freight allocation plans are to be determined. This problem is formulated as a mixed integer linear programming model. An iterative scheduling approach is designed to solve the model. A numerical example based on Union-Pearson Express in Toronto is introduced to demonstrate the efficiency of the proposed methods.
Electric Vehicle Traveling Salesman Problem with Drone
Tengkuo Zhu, University of Texas, AustinShow Abstract View Presentation
Stephen Boyles, University of Texas, Austin
Avinash Unnikrishnan, Portland State University
The idea of deploying electric vehicle and unmanned aerial vehicles, also known as drones, to perform "last-mile" delivery in logistics operations has attracted increasing attention in the past few years. In this paper, an Electric Vehicle Travelling Salesman Problem with Drone (EVTSPD) is formulated as a mixed integer linear program to aid logistic organizations with a new method of delivering parcels which can extend the driving range of both vehicles and decrease the operation cost. An iterative solution heuristic is also developed. Results of numerical experiments show that the heuristic is much more efficient than solving the problem via a standard solver. Incorporating UAVs into EV based routing were found to reduce average delivery times by up to 40% for the instances tested.
Analytic Approximations of Realistic Tour Distances: Case Studies for Deliveries by Robots and Drones
Youngmin Choi, University of Maryland, College ParkShow Abstract View Presentation
Paul Schonfeld, University of Maryland, College Park
Due to the large computational effort needed to solve a Travelling Salesman Problem (TSP), researchers have developed approximations of the average length of TSP tours . These approximation models have been widely applied in transportation planning for optimizing large-scale system problems. In response to growing interest in last-mile deliveries by vehicles with limited carrying capacities, this study focuses on a TSP tour approximation with a realistic number of points that can be visited per tour by such vehicles. The proposed model accounts for various operating conditions, including distance measures, shapes of service area, and locations of distribution center. The model is applied to analyze deliveries by ground robotic vehicles (robots), unmanned aerial vehicles (drones), and conventional trucks. Deliveries by robots and drones have lower total cost than by trucks for our baseline inputs. Sensitivity analyses are designed to explore how system outputs vary with changes in the baseline. Drones can be a cheaper delivery option than ground robots if energy cost is near the baseline, but the difference in total costs diminishes as that cost increases. At high value of customer time spent waiting for deliveries, drones can be the most cost-effective option. Delivery vehicles with larger carrying capacity may be favored for service areas that are larger or have higher demand densities.
Deep Learning–Based Crowdshipping Delivery Production Forecasting
Hui Shen, University of Illinois, ChicagoShow Abstract View Presentation
Jane Lin, University of Illinois, Chicago
Crowd-shipping (CS) is an emergingdelivery service that occasional drivers sign up to deliver goods using empty space in their vehicles. Much of the demand and supply of CS is under studied. In light of that, this study investigates a short-term CS delivery requests forecasting problem by utilizing empirical data from a real-world CS platform and two deep learning methods, namely the long short-term memory neural network (LSTM NN) and the bidirectional long short-term memory neural network (BDLSTM NN). City of Atlanta, Georgia is chosen as a case study. The spatial and temporal analysis of the distribution of delivery requests by zip code finds significant variability zip code by zip code, in both delivery production and attraction. The advanced deep learning based CS delivery production forecasting models show that the CS delivery production is affected not only by the past number of delivery requests but also weather conditions and time period of a day. These two deep learning models are compared with other commonly used statistical and machine learning models and found to outperform other benchmark methods in terms of RMSE. Generally speaking BDLSTM NN out performed LSTM NN slightly.