Integrative Framework for the Predictive
Evaluation of Traffic Control Strategies

EPSRC: GR/K 16593

Researchers: Evtim Peytchev & Andrew Argile

Introduction

During the last two decades significant research effort has been invested into the development of various computer assisted traffic control schemes. The expansion of these control schemes reflected the changes in optimization criteria that evolved gradually in line with the accumulated operational experience. Early traffic control systems essentially performed static optimization of the maximum bandwidth of the "green waves" on specific routes in the city. This worked well on lightly or moderately loaded signalized arterials with few vehicles turning in from side roads. However, the optimization of a more complex city traffic required taking into account other performance indices such as the length of queues on traffic lights or the number of vehicle stops during a journey. More recent control systems use the on-line traffic flow data to modify the split/cycle/offset timing of the road junction lights in response to random fluctuations of traffic intensities. The projected developments of the traffic management systems of the 90's incorporate the inclusion of priority levels for various groups of users, road pricing, in-car route guidance systems, etc. As the control strategies grow in complexity, requiring expensive real-time measurements to maximize the road capacity, there is a need for the assessment of their effectiveness.

While the test scenarios for the field trials of the early control systems, which were concerned with the average levels of performance, were relatively easy to design, the evaluation of the new control schemes is a much more demanding task since it requires extensive probabilistic studies that relate the controls and the random fluctuations of traffic flows.

This research is focused on the development of a flexible and portable real-time system for non-intrusive evaluation of urban traffic control schemes. An important requirement "not to disrupt the operation" of the actual control system while "offering the flexibility of modular augmentation" of progressively more sophisticated simulators, directed this research towards the development of a distributed computing environment.

Also this research recognizes that the simulation results can only be correct in a statistical sense. Consequently, since the simulations are to be used in the process of deriving control decisions, this research aims at developing a simulator that quantifies the confidence limits on the results it produces. It is, it evaluates to which extent the discrepancy between the assumed and the actual traffic volumes and the random variation of drivers decisions, affects the accuracy of calculated traffic flows, journey times, average queue length etc.

Project Progress

• Integrative framework for the predictive evaluation of traffic control strategies
  A conceptual design of the framework has been accomplished. As illustrated on the structure diagram, the major functional modules of predictive simulation and control have been separated from each other and from the actual traffic control system by means of a Distributed Shared Memory interface. Consequently the overhead imposed on the evaluated control system is minimal and it involves the maintenance of the communication line only.
  
  
• Probabilistic ADaptive SImulation Model (PADSIM)
  The PADSIM model has been developed in order to combine the realism of microsimulations with the efficiency of macrosimulations without incurring the dissadvantages of the respective approaches. The PADSIM simulator achieves high computational performance by means of macroscopic modelling and it retains realism through callibrating the macrosimulations using concurrent microsimulations in selected network nodes (this can be considered to be logically equivalent to taking measurements).
  
  • Sample screen illustrating the operation of the microsimulator is provided here.
    The macrosimulation module is currently being developed.
    
  • A paper on the design of PADSIM has been presented at the European Simulation Multiconference ESM'95.
  
  
• Distributed Shared Memory (DSM) system
  Communication between the various software modules executing on distributed computing nodes has been achieved by implementing a virtual distributed shared memory (DSM) system. This approach offers implementation independent interface to the applications programmer. The operation of the DSM system has been successfuly evaluated in trials over Internet involving heterogenous computing nodes at the Nottingham Trent University and Helsinki University of Technology.
  Two application tasks and a shared memory manager task have been spawned and performed read/write operations on shared memory. The Dummy SCOOT (data provider) task was designed to mimick the operation of the real SCOOT system and to write historical traffic mesasurements to shared memory (Memory Manager task). This measurement data was subsequently retrieved by a generic Data Analysis (data consumer) task which, in the test scenario, was the measurement visualisation task, but it could also be a PADSIM simulator.
  
  
• Traffic Network Initialization and Visualisation Module
  A versatile X11-based graphics library has been developed and used in implementing a network visualisation module to be used by PADSIM simulators. The module uses the traffic network description format which is compatible with the City Traffic Map Editor (TEDI), developed by the Laboratorio de Investigacion Operativa y Simulacion at the Universitat Politecnica de Catalunya (UPC).
  
  Various urban traffic network maps were created using the TEDI editor and were imported for use by PADSIM modules.

Industrial Collaborators

• Nottingham Traffic Control Centre
• Leicester Area Traffic Control

Related Developments

The feasibility the integration of an alternative powerfull microsimulation module HUTSIM, developed at Helsinki University of Technology, with the PADSIM macrosimulator, is being investigated as an extension of the current research. This collaborative research between the NTU and HUT is fully funded by the Finnish Academy of Sciences.

---

Copyright (c) 1995 RTTS
Last update: 20/12/95

---

Simulation Modelling - home page