Research Projects Supported by HKU's High Performance Computing Facilities
¡@
Researcher:
Miss Wing-man Cheung, Department of Civil Engineering
Project Title:

Dynamic Traffic Assignment for Congested Highway Network
Project Description:
This project incorporates traffic assignment into origin-destination (O-D) matrix estimation problem to form a dynamic traffic assignment (DTA)/O-D estimation problem in order to cater for congestion effect. DTA problems and O-D estimation problems are formulated as optimization problems. This study first investigates the use of adaptive genetic algorithms (AGAs) for solving the DTA/O-D problem because AGAs are simple but powerful techniques for solving optimization problems with non-convex objective functions. They are also parallel in nature so that computation time can be shortened with parallel computing.
Project Duration:
2 years
Project Significance:
Highway transport problems, for instance, congestion, accidents and pollution, can be mitigated by better transport planning. A good transportation planning study requires the availability of a reliable traffic demand forecasting model and a reliable traffic flow model. Advanced measures to mitigate the transport problems make use of the Intelligent Transport Systems (ITS). ITS works by measuring real-time traffic flow, detecting accidents instantly and providing up-to-date traffic information to road users. It requires improvements in the technology for in-vehicle communication, as well as techniques for modeling traffic flows and predicting traffic conditions. DTA predicts traffic flows in a congested highway network by modeling dynamic route choice behavior of road users upon receiving full or partial real time transportation information. The implementation of parallel computation to solve DTA problems quickly enables simulating real-time traffic condition accurately and thus the successful functioning of ITS.
- Back to top- / Contents
Results Achieved:

Experimental results obtained from the research shows that the use of AGA instead of conventional methods to solve DTA/O-D estimation problems can achieve a more accurate solution. Since AGA can be solved through parallel computation while the conventional methods are sequential in nature, computation time consumed by AGA is much lower.  

The graph below also shows that the more computer processors are used to solve AGA, the higher speedup ratio can be reached. In addition, the actual speedup is very similar to the actual speedup which means that parallel computation is very efficient in solving DTA/O-D estimation problems by means of AGA.

Remarks on the Use of High Performance Computing Cluster:
Only with the powerful HPC Cluster can this research investigate the advantages of using AGAs in solving DTA/O-D estimation problems. It also enables the study of efficiency of implementing parallel computation in AGAs. As revealed from the promising results of the research, implementation of parallel computation to solve DTA/O-D estimations problems quickly can enable simulating real-time traffic condition accurately and thus the successful functioning of ITS.
- Back to top- / Contents