METRANS UTC

Optimum Routing of Freight in Urban Environments under Normal Operations and Disruptions using a Co-simulation Optimization Control Approach

Project Number

15-12

Project Summary

The complexity and dynamics of the road and rail networks that are also shared by passengers  together with the unpredictability of the effect of incidents, disruptions and demand, in temporal and spacial  coordinates makes the scheduling and optimum routing of freight a very challenging task despite recent  advances in information technologies. Estimating travel times in an urban road network environment is a real  challenge especially during incidents and other disruptive events. Current practices rely on historical data and  limited available real time information in order to make routing decisions that minimize a certain cost  objective which in the case of road network is usually travel time. Incidents, disruptions, changes in demand,  planned and unpredictable events may change the historical patterns of traffic, rendering decisions ineffective  leading to imbalances in capacity across time and space coordinates.

Project Status

Complete

Project Report

Project Brief

Year

2015

Topic Area

Integrated Freight and Passenger Systems

P.I. Name & Address

Professor of Electrical Engineering Systems, Ming Hsieh Department of Electrical Engineering; USC Viterbi School of Engineering
University of Southern California
3740 McClintock Avenue
Hughes Aircraft Electrical Engineering Center (EEB) 200B
Los Angeles, CA 90089-2562
United States
ioannou@usc.edu

Funding Source

U.S. Department of Transporatation

Total Project Cost

$100,000

Agency ID or Contract Number

Grant No: DTRT13-G-UTC57

Start and End Dates

7/1/2015 - 6/30/2016

Brief Description of
Research Project

The complexity and dynamics of the road and rail networks that are also shared by passengers  together with the unpredictability of the effect of incidents, disruptions and demand, in temporal and spacial  coordinates makes the scheduling and optimum routing of freight a very challenging task despite recent  advances in information technologies. Estimating travel times in an urban road network environment is a real  challenge especially during incidents and other disruptive events. Current practices rely on historical data and  limited available real time information in order to make routing decisions that minimize a certain cost  objective which in the case of road network is usually travel time. Incidents, disruptions, changes in demand,  planned and unpredictable events may change the historical patterns of traffic, rendering decisions ineffective  leading to imbalances in capacity across time and space coordinates.
 
The purpose of this project is to exploit the availability of powerful computational and software tools  together with advances in optimization and feedback control for dynamical systems in order to come up with a  methodology that can lead to more efficient decisions in freight scheduling and routing. The method relies on  the use of real time simulation models, for predicting travel times, traffic flows, fuel consumption and  pollution by going beyond to what can be achieved today based on historical and limited real time data. The  simulation models are used in a feedback loop with an optimization model and a load-balancing controller. The  simulation models receive historical and streaming data and are able to automatically reconfigure themselves  to simulate the subsequent effects of incidents and disruptions. They can be used to estimate costs (travel time, fuel consumption, pollution etc.) along different links in the network by fast-forwarding. In addition they can  be used to test different decision scenarios before reaching a final decision. The predicted states of the network  can be used to generate cost estimates along possible routes, which in turn can be used by an optimizer to  calculate the optimum route with respect to space and time. In many cases however routing decisions and  expected future demand may easily disturb the states of the network and change the initially estimated costs  leading to an unbalanced network load in space and/or time. A load balancing controller exercises the  simulation model and tests different load distributions along the possible routes. The approach leads to an  iterative feedback process with the objective of reducing the value of the cost index further till a stopping  criterion is met in which case the final decision is applied to the real system. The objective of this project is  focused on the effectiveness of the methodology as a tool for freight scheduling and routing in a complex  urban environment. We like to investigate whether the use of simulation models operating in real time with  optimization and automated control techniques can provide much better decisions than existing approaches that  rely on past data and limited real time information. Issues such as speed of computations, scalability,  convergence, ability for fast reconfiguration during incidents and disruptions are important research problems.
 
While most of the work will concentrate on developing and analyzing the main components of the proposed  co-simulation optimization control approach all simulations, testing and evaluations will be carried out by  using a validated microscopic simulation model of part of the rail and road network in the Los Angeles and  Long Beach area that includes the two major ports. The deliverables of the project will include the  demonstration of the co-simulation optimization control approach and its benefits in upgrading current  practices to a new level that takes full advantage of available technologies that include computations and  automation of decisions in addition to information technologies. Our experience with traffic models indicates  that even though duplicating the real world is an impossible task, simulation models provide much better  information than static or equilibrium models and therefore their use in managing freight flows and traffic in  general offer a strong potential for dramatic improvements if properly used in combination with optimization  and control techniques. This project is aimed to demonstrate this potential as a main deliverable.