Agent-Based Simulation Enhanced with Big Data and Artificial Intelligence Methods to Simulate and Optimize Traffic Flow

Traffic systems are characterized by a number of features that make them hard to analyze, control and optimize. Due to the increasing traffic demand, even modern societies with well-planned road management systems and infrastructures for transportation still confront problems such as congestion, delays and even accidents. Constructing new roads could be one of the solutions for handling the traffic congestion problem but often, the extension of the actual infrastructure is very hard due to the large cost or impossible because of the lack of space. Furthermore, complex urban transport systems are composed of many interacting parts connected in complex ways where the laws guiding the overall system behavior are unknown. In these scenarios, traffic simulation can provide a precious tool to represent, analyze and solve traffic problems. It allows to identify and manage traffic patterns in large scale multi-modal transportation systems by making more efficient use of the existing infrastructure.

Agent-based simulation (ABS) is relatively a new approach to model complex traffic systems by dividing them into a set of interacting autonomous agents that learn the global dynamics of the traffic system from the interactions of many individual behaviors. Our ultimate goal in this research is to integrate artificial intelligence (AI) technologies with agent-based simulation to realize the following three challenging objectives:

Develop a realistic simulation model for Al-Ain city calibrated using actual traffic Big Data that exhibits and demonstrates a realistic pattern of traffic flow, thereby allowing evaluating and visualizing network performance.
Using AI-based approaches namely Multi-Objective Metaheuristics, Reinforcement and Deep learning to optimize network flow by optimized signal timing and route choices

Our approach integrates multi-objective optimization (e.g., NSGAII ), data mining (e.g., deep and reinforcement learning) to the multi-Agent traffic Simulation in order to enhance the efficiency of traffic solution that are in nature dominated by multi-criteria decision making, prediction and optimization. The proposed multi-agent traffic simulation model will be calibrated using real traffic Big data and optimized using AI techniques. A realistic pattern of traffic flow, thereby will allow analyzing, making cost-effective traffic-control strategies for Al-Ain traffic network.