Research
Mathematical optimization and simulation for transport, logistics and public safety.
I develop quantitative methods for decisions in complex, uncertain systems and test them on large real-world datasets, usually with an operator or a public authority supplying the data.
Large events push transport systems into states a regular timetable never sees. For the FIFA World Cup 2022 in Qatar I built an agent-based transport demand simulation covering up to 800,000 agents per simulated day, including residents, expatriates, foreign visitors and the event workforce. We ran it before the tournament to find the bottlenecks in the metro network, and the mitigations went into the operational plan. The same question runs through the rest of this work: how do you keep a crowd moving when demand exceeds capacity? At the Hajj, reassigning service offices to paths cuts the peak load at the worst bottleneck by 47 percent against the assignment used in 2018, certified within 0.18 percent of the best achievable. That figure comes from a planning model and has not been implemented.
Dkaidik, U., Haase, K., Koch, M., & Vlćek, T. (2026). Mitigating Crowd Congestion at the Hajj: Lexicographic Minimax Path Assignment for the Mina Tent City. Operations Research Proceedings 2026 (accepted, forthcoming).
Dkaidik, U., Haase, K., Koch, M., & Vlćek, T. (2026). Spatial Planning of the Mina Tent City: Crowd-Safe and Administrable Camp Assignments for the Hajj. Working paper.
Vlćek, T., Dkaidik, U., Haase, K., Koch, M., Weygandt, A., Pagels, L., & Pape, J. (2025). Network-Wide FIFO Inflow Control for Oversaturated Urban Rail Transit: A Cumulative-Count Optimization Framework. Working paper. DOI GitHub
Bonz, J., Craze, B., Haase, K., Halligan, M., Koch, M., & Vlćek, T. (2024). Controlling the Transport Demand of the FIFA World Cup 2022 in Qatar. SSRN preprint. DOI
The boundaries of a service district decide who waits longest. For police service district planning I combined a districting optimization with discrete-event simulation and found that relocating stations cuts simulated response time by up to 14.5 percent, while redrawing the borders alone gives 6.0 percent and a third station in the Belgian case gives 8.7 percent. The follow-up work asks what patrol car allocation does to response time once the districts are fixed. The same structure turns up well outside public safety: in sales territory design I solved instances with 2,020 locations to proven optimality in about 141 seconds, on synthetic data where the previous state of the art could not prove optimality in two hours.
Cors, T., Fliedner, M., Haase, K., & Vlćek, T. (2025). Simulative assessment of patrol car allocation and response time. European Journal of Operational Research, 324(1), 290–307. DOI
Vlćek, T., Haase, K., Fliedner, M., & Cors, T. (2024). Police service district planning. OR Spectrum, 46, 1029–1061. DOI GitHub
Cors, T., Vlćek, T., Voß, S., & Jovanovic, R. (2023). Fixed Set Search Applied to the Territory Design Problem. Metaheuristics, 322–334. DOI
Vlćek, T. (2023). The Sales Force Deployment Problem for Teams of Sales Representatives Within Sales Territories. Operations Research Proceedings 2022, 121–127. DOI GitHub
This work usually starts with an operator who already has a plan that works and wants a better one. For nine mobile library areas in Germany, covering 1,141 locations and 1,278 visits, a periodic routing matheuristic cuts driven distance by 13 to 33 percent per area against the manually built plans, with a median of 18.6 percent. The libraries use the results to improve their tours. For a rural German regional transport operator, an on-demand line-based service reaches 3.3 percentage points more service level than the commercial dispatcher they had deployed, on the same fleet, at 0.06 seconds per booking event across 36,414 bookings, six depots and 76 days. The third strand is warehousing, where I work on allocating SKUs across warehouses so that fewer orders have to be split into several parcels.
Schulz, A. & Vlćek, T. (2026). Evaluation of On-Demand Line-Based Bus Services. Under review at Transportation Science. GitHub
Vlćek, T. & Haase, K. (2026). The Periodic Vehicle Routing Problem with Balanced Visit Spacing: A Matheuristic and a Case Study in Mobile Library Routing. Working paper.
Vlćek, T. & Voigt, G. (2026). Optimizing SKU-Warehouse Allocations to Minimize Split Parcels in E-Commerce Environments. Working paper. GitHub
Where this is going
The direction I want to take this next is the intersection with AI: agent-based simulation in which language models supply the agents' behavior instead of hand-coded decision rules, and decision and learning systems that people can actually work with.