Every traveller, simulated.

The capacity planner offers the morning-peak car trips of every prospective home as agents that can either travel or opt out. Agents compete for road space. When part of the network approaches gridlock, measured by its macroscopic fundamental diagram, travelling there becomes costly and the marginal trips drop out, and each block keeps homes in proportion to its home trips that get through.

It runs in two modes. The over-allocation check keeps the design's allocation and asks where it is over-allocated and how many homes per block survive. The planner's answer is a starting point: full-day simulations then scale its homes up or down until the busiest part of the network peaks just below its critical load. The sensitivity map lets the same homes move between blocks and shows where the network has most room. It is not a plan.

The method grew out of our flood evacuation work (below) and was rebuilt on current MATSim for urban growth. The interactive demo runs it on the Tanjong Pagar waterfront scenario, designed by Tanvi Maheshwari at the Future Cities Laboratory for her ETH Zurich doctoral thesis and simulated by Pieter in Sketch MATSim. It is also the test case of our joint papers in the Journal of Urban Design and Philosophical Transactions of the Royal Society A.

In a floodplain, every new dwelling adds a vehicle to an evacuation that already depends on a few roads and bridges. The Flood Capacity Planner asks where additional development can go without materially increasing risk to life.

Candidate development enters the evacuation model as surplus demand. Each surplus agent chooses between evacuating through a subsector or not being built, and routing with decongestion-based replanning finds the allocations the network can support across an ensemble of flood events. The result is validated against the official evacuation model.

Planners load a scenario, place demand, simulate, validate, and compare risk to life against the baseline on a map. The planner works alongside the evacuation model and makes it more useful for decisions.

The Hawkesbury-Nepean Valley in western Sydney is one of the most flood-exposed places in New South Wales, and its roads out are few. The Flood Evacuation Model (FEM) simulates the evacuation of every vehicle in the valley against ensembles of flood events, up to the probable maximum flood. Rising water closes roads as it would in reality, and agents learn routes to safe destinations under the evacuation plan.

FEM2 began with Kai Nagel at TU Berlin in late 2017 and was developed with URaP International and CSIRO Data61. Pieter is the main architect and developer of its core engine. The model informs flood risk management planning for the valley; Infrastructure NSW describes its use in a public 2023 technical report.

At ETH Zurich’s Future Cities Laboratory in Singapore, Pieter built the demand side of the first MATSim model of Singapore: a synthetic population of about five million agents, each located at an individual building. Fare-card data then made it possible to simulate public transport at the level of individual buses and passengers, the subject of Pieter’s ETH doctorate, Data-Driven Transit Simulation.

As project leader of the Engaging Mobility team, Pieter took the work into urban design: autonomous on-demand transit and parking for the Tanjong Pagar waterfront with NUS and SMART, and Sketch MATSim, developed with Sergio Ordóñez, which lets designers test street and land-use layouts against simulated travel. Tanvi Maheshwari designed the Tanjong Pagar waterfront and a series of design experiments for her doctoral research; Pieter simulated them in Sketch MATSim, and Tanvi read the results as an urban designer. That pairing of urban design and transport simulation led to two joint papers.