Abhilasha Saroj

Abhilasha Saroj, Shaked Regev, Guanhao Xu et al.

Traffic simulation and digital-twin calibration is a challenging optimization problem with a limited simulation budget. Each trial requires an expensive simulation run, and the relationship between calibration inputs and model error is often nonconvex, and noisy. The problem becomes more difficult as the number of calibration parameters increases. We compare a commonly used automatic calibration method, a genetic algorithm (GA), with Bayesian optimization methods (BOMs): classical Bayesian optimization (BO), Trust-Region BO (TuRBO), Multi-TuRBO, and a proposed Memory-Guided TuRBO (MG-TuRBO) method. We compare performance on 2 real-world traffic simulation calibration problems with 14 and 84 decision variables, representing lower- and higher-dimensional (14D and 84D) settings. For BOMs, we study two acquisition strategies, Thompson sampling and a novel adaptive strategy. We evaluate performance using final calibration quality, convergence behavior, and consistency across runs. The results show that BOMs reach good calibration targets much faster than GA in the lower-D problem. MG-TuRBO performs comparably in our 14D setting, it demonstrates noticeable advantages in the 84D problem, particularly when paired with our adaptive strategy. Our results suggest that MG-TuRBO is especially useful for high-D traffic simulation calibration and potentially for high-D problems in general.

Somdut Roy, Michael Hunter, Abhilasha Saroj et al.

Emergency response vehicles (ERVs), such as fire trucks, operate to save lives and mitigate property damage. Emergency vehicle preemption (EVP) is typically implemented to provide the right-of-way to ERVs by giving green signals as they approach signalized intersections along their routes. EVP operations are usually optimized to minimize ERV delay. This study seeks to reduce delay experienced by other vehicles in the network while keeping ERV travel time near its optimum. A machine learning-based EVP strategy, termed MLEVP, is developed to determine EVP trigger times at multiple downstream intersections using real-time sensor data, including vehicle detections, signal indications, and ERV location. MLEVP proactively clears downstream traffic queues to reduce ERV response time while limiting delay on conflicting traffic movements. In the case study, MLEVP is developed using a calibrated microscopic simulation of a signalized corridor testbed in PTV Vissim. The EVP problem is formulated as a regression problem and solved using machine learning models trained on data generated from the simulation. Results demonstrate that the proposed algorithm can produce near-optimal ERV travel times while minimizing impacts on conflicting traffic.