Real-Time Predictive Control Strategy Optimization

TL;DR

The paper tackles real-time traffic management on large-scale networks by integrating predictive control of network tolling with consistent traveler guidance within a Dynamic Traffic Assignment framework. It introduces a rolling-horizon architecture built around DynaMIT2.0, where state estimation, state prediction, and strategy optimization operate iteratively to produce both toll schedules and feedback guidance that reflect anticipated network conditions, converging via the method of successive averages. To solve the resulting nonlinear, simulation-based optimization, the authors deploy a highly parallel real-coded Genetic Algorithm (NSGA-II) with Batch-Wise evaluation and a Master-Slave GNU Parallel implementation, achieving sub-5-minute computation per cycle. Experiments on Singapore’s expressways show predictive tolling improves network-wide travel times by up to 8–9% over no-toll and static-toll benchmarks, with robust performance across demand levels and significant gains demonstrated under real-time, closed-loop evaluation. This framework offers a scalable, data-driven approach for real-time congestion pricing and traveler-information systems with practical implications for urban traffic management.

Abstract

Traffic congestion has lead to an increasing emphasis on management measures for a more efficient utilization of existing infrastructure. In this context, this paper proposes a novel framework that integrates real-time optimization of control strategies (tolls, ramp metering rates, etc.) with guidance generation using predicted network states for Dynamic Traffic Assignment systems. The efficacy of the framework is demonstrated through a fixed demand dynamic toll optimization problem which is formulated as a non-linear program to minimize predicted network travel times. A scalable efficient genetic algorithm is applied to solve this problem that exploits parallel computing. Experiments using a closed-loop approach are conducted on a large scale road network in Singapore to investigate the performance of the proposed methodology. The results indicate significant improvements in network wide travel time of up to 9% with real-time computational performance.

Figures (13)

• Figure 1: Framework for Integrated Guidance Generation and Control Strategy Optimization
• Figure 2: Illustration of the rolling horizon approach for toll optimization
• Figure 3: Genetic Algorithm with parallel evaluation of population using parallel & distributed computing techniques.
• Figure 4: Closed-Loop Framework
• Figure 5: Network of Expressways and Major Arterials in Singapore