Smart Routes in Motion
AI driven routing helps logistics adapt in real time, reducing delays and improving efficiency for cities, businesses, and on demand services facing uncertainty and dynamic demand.
Dawei Chen, PhD researcher in the Operations Planning, Accounting & Control group, earned his doctorate at Eindhoven University of Technology on June 2, 2026.
Chen investigated how artificial intelligence can improve vehicle routing decisions in uncertain and fast changing environments.
Real challenge
Every day, logistics providers deal with unpredictable travel times, sudden traffic congestion, and continuously arriving customer requests. Traditional planning methods fix routes in advance, making them difficult to adjust once vehicles are on the road. This leads to inefficiencies that affect delivery companies, ride sharing services, and food platforms, especially in busy urban areas where conditions can change within minutes.
Living routes
Chen shows that routes do not need to be fixed. His research demonstrates how algorithms can continuously adapt routes based on real time traffic information. Instead of planning once at departure, the system learns to revise decisions during the trip, improving responsiveness. This creates opportunities for logistics companies to better handle delays and disruptions without constant manual intervention.
Moving targets
The research becomes more complex when customer orders appear dynamically. In this situation, vehicles must decide not only where to go, but also when to leave and which orders to prioritize. Chen developed a reinforcement learning approach that helps systems learn from these decisions over time. The challenge lies in the uncertainty between orders, which makes it difficult for conventional models to predict travel times and plan efficiently.
Food flows
A practical application explored in the thesis is online food delivery. Here, timing is critical because meals must remain fresh while customers expect fast service. The developed AI method determines when to pick up orders from restaurants and how to combine multiple deliveries into efficient routes. Using data based on real road networks and actual order patterns, the system demonstrated clear benefits for platforms and their users.
Clear results
The results show that Chen鈥檚 approach can serve more customers while reducing average delays from 17 minutes to 7 minutes. This has direct implications for businesses aiming to improve service levels while reducing operational costs. For society, this can translate into less congestion, fewer emissions, and more reliable delivery services that fit the pace of modern life.
Wider lens
The work within the Operations Planning, Accounting & Control group of the department of Industrial Engineering and Innovation Sciences, highlights how tailored AI solutions can outperform more general systems in complex, uncertain settings. Chen鈥檚 findings suggest that specialized learning approaches can form the basis for addressing even more complicated problems, such as integrating multiple sources of uncertainty across supply chains and urban mobility systems.
Dawei Chen defended his thesis on June 2, 2026.
Title of the thesis:
Supervisors: Tom van Woensel, Christina Imdahl.