Proposal: Scalable kernel-based simulation of travel demand with dynamic network feedback Dynamic traffic modelling requires disaggregate travel demand that captures individual activity schedules and their interaction with the transport network. Recent work has introduced a kernel-based framework for synthesizing travel demand by transferring behaviour from travel survey respondents to agents in a synthetic population. The method combines similarity kernels over socio-demographics, travel schedules, and activity-location feasibility within a Metropolis-Hastings sampling scheme to generate daily travel plans. Initial experiments show that the approach reproduces marginal travel patterns while preserving individual variability. Scaling the method to larger populations and incorporating network feedback requires substantial computational resources. This project aims to extend the framework using high-performance computing (HPC) infrastructure. The objective is to simulate synthetic travel demand for large urban regions, potentially up to about one million agents, while incorporating network conditions and feedback mechanisms, enabling systematic scenario analysis. The first objective is scaling the demand synthesis procedure. The method evaluates similarity between synthetic agents and survey respondents across several behavioural dimensions. As the number of agents and survey records increases, the number of pairwise kernel evaluations grows rapidly. HPC resources will allow efficient evaluation of these kernels and make it possible to incorporate additional behavioural and spatial dimensions. The second objective is integration of network conditions in the synthesis process through congestion feedback mechanisms. Travel plans determine traffic flows, which in turn influence travel times and accessibility on the network. The project will incorporate travel times obtained from dynamic traffic simulation or network assignment models. These travel times will enter the sampling procedure through schedule characteristics and feasibility constraints. Iteratively updating travel plans using network conditions allows the synthesized demand to remain consistent with congestion patterns and infrastructure constraints. The third objective is to explore parallelization of the sampling algorithm. The Metropolis-Hastings framework modifies subsets of agents at each iteration and repeatedly evaluates kernel similarities and likelihood ratios. Several parts of this procedure can be parallelized, including kernel evaluation across agents, proposal evaluation, and running multiple Markov chains. HPC systems also allow multiple policy scenarios to be simulated in parallel, supporting analysis of infrastructure investments, land-use changes, and demand management measures. The result will be a scalable framework for generating synthetic travel demand that remains consistent with both survey observations and network conditions. This will support dynamic traffic simulations and policy analysis in large-scale transport modelling applications.