Ph.D. - Computer Science
University of Ottawa
Rym Zalila-Wenkstern
Professor - Computer Science
Director, Center for Smart Mobility
Director, Smart Cities Applied Research Lab
Professional Preparation
Doctorat de Specialite - Computer Science
University of Tunis
University of Tunis
Engineering Degree - Computer Science-Software Engineering
University of Tunis
University of Tunis
Pre-Engineering - Mathematics, Physics and Chemistry
University of Tunis
University of Tunis
Baccalaureat - Mathematics and Physics
Academie de Paris
Academie de Paris
Research Areas
Research Interests
- Intelligent Transportation Systems
- Autonomous Vehicle Systems
- AutonomousTraffic Control Systems
- Multi-Agent Systems
- Large-scale Multi-Agent-Based Simulation Systems
Awards
Best System/Demo Award - International Conference on Autonomous Agents and Multi-Agent Systems [2020]
Smart 50 Award - Smart Cities Connect Foundation [2019]
Finalist, Tech Titans Award (innovator category) - Tech Titans [2019]
Hometown Technology Hero - City of Richardson [2019]
Outstanding Service to CS Department - UT Dallas [2017]
Outstanding Faculty Teaching Award - UT Dallas [2015]
Outstanding Service to the CS Department - UT Dallas [2014]
Best System/Demo Award - International Conference on Autonomous Agents and Multi-Agent Systems [2013]
Best Overall Paper Award - Simulation Multi-Conference [2013]
Best Paper Award - Agent-Directed Simulation Symposium [2013]
Projects
DALI - An Autonomous, Collaborative Traffic Control System
DALI is a pioneering agent-based, collaborative traffic control system aimed at alleviating congestion and enhancing the driving experience. Its objective is to infuse existing traffic signal timing systems with autonomy and intelligence, without the need for costly changes or upgrades. This is achieved by integrating an intelligent software agent into each existing intersection controller, effectively serving as the controller's "brain." These agents analyze traffic data, engage in direct communication, and collaboratively implement real-time timing strategies to enhance traffic flow.As a plug-in, AI-driven software solution, DALI has the capability to scale seamlessly to cities of any size. Its successful implementation in the City of Richardson, Texas, yielded initial results showcasing a substantial 40% reduction in traffic delays. Presently, the system is undergoing broader deployment within the City of Richardson. If this endeavor proves successful, DALI would stand as the first fully collaborative, agent-based traffic control system to achieve large-scale deployment within the United States.
Autonomous Vehicle Systems
This project focuses on developing advanced cooperative planning algorithms for Connected and Autonomous Vehicles (CAVs) operating in unpredictable environments. Leading AI-based methods generally frame the CAV planning problem as a Multi-agent Markov Decision Process (MMDP), solved using Monte Carlo Tree Search (MCTS). However, when applied to scenarios with a high count of CAVs, MCTS encounters scalability issues due to the exponential growth of the branching factor with CAV numbers. In response, we have introduced three MCTS-based algorithms that enhance performance and scalability. These algorithms were integrated and thoroughly assessed within the MATISSE simulator. The experimental findings demonstrate their superiority over state-of-the-art optimization-based and Reinforcement Learning (RL)-based algorithms.MATISSE - A Large-Scale Multi-Agent-Based Traffic Simulation System
MATISSE is the only microscopic traffic simulator built from the ground up as a multi-agent system. It offers a range of features for simulating Intelligent Transportation Systems (ITS), including various virtual agent types (e.g., virtual drivers, connected vehicles, autonomous vehicles), simulated sensor-based perception mechanisms for both virtual autonomous vehicles (e.g., virtual LiDAR, virtual radar) and drivers (e.g., vision, hearing), virtual smart and conventional intersection controllers, simulated I2I, V2V, and V2I communications, real-time modification of virtual vehicle and driver properties without interrupting the simulation, spontaneous generation of unscripted accidents, automatic conversion of Open Street Map road networks, and automatic generation of missing information (e.g., lane numbers, traffic light locations). Additionally, MATISSE supports hybrid simulations that incorporate real-time field data. No current traffic simulator provides such a comprehensive set of features.DIVAs - A Large-Scale Multi-Agent Simulation Framework
Developing Multi-Agent-Based Simulators (MABS) with open environments is a complex task. It involves defining distinct models for agents and the environment, enhancing agent architectures with efficient perception mechanisms, decentralizing the environment architecture, and defining accurate agent-environment interaction protocols.Dr. Wenkstern's research team has introduced a novel model that formalizes interaction protocols between virtual agents and the environment. This model enables the simultaneous execution of up to 4,000 virtual agents on a single desktop. In a simulation cycle lasting less than 160 milliseconds, two steps occur: 1) Virtual agents perceive their surroundings using sensors, combine data, communicate, deliberate, and decide on actions; and 2) The environment combines agents' intended actions, ensures compliance with physical laws, determines outcomes, updates its state, shares the state with agents, and accommodates user-triggered events. The model handles the most challenging open and spatial environments, previously unaddressed in published work. It serves as the core component of DIVAs, a large-scale multiagent simulation framework.