Gabriel Gomes, of UC Berkeley PATH, presented Open Traffic Models: A Platform for Large-scale Hybrid Traffic Simulation at the ITS Seminar April 19, 2019 at 4 p.m. in 290 Hearst Memorial Mining Building.
Abstract
Dr. Gomes is a research engineer at ITS, and also a lecturer with the Mechanical Engineering department. He came to Berkeley in 1999 as a graduate student in control theory, and focused on problems in traffic modeling and control. Following his graduate work, he stayed on at CCIT where he worked on the NGSIM project, and later with PATH where he has been part of several projects including TOPL and Connected Corridors. Recently Dr. Gomes has been working more closely with the Mechanical Engineering department to develop and supervise capstone projects for the Master of Engineering program. He has also worked with LBNL to develop techniques for large scale traffic simulation in High Performance Computing environment, and we will hear about some of that work today.
Presenter
Although traffic modeling and simulation have been an object of study for many decades, there is still no model that can be considered "best" in all situations. Rather, it is recognized that different models are better suited for different scenarios, and in some cases a combination of models is desirable. This observation has lead to the concept of hybrid simulation, in which traffic simulators of different types are stitched together at an interface. Research into this concept has thus far focused on the combination of particular pairs of models; e.g. CTM with Newell's car following model, software A with software B, etc. In this talk, Dr. Gomes will delve into the problem of interfacing a pair of arbitrary macroscopic, mesoscopic, or micrscopic models. The concepts that will be presented have been implemented in the Open Traffic Models framework, or OTM. OTM provides a means for creating and combining light-weight plugins for traffic modeling and control. The two main components of OTM are a) a generic data schema for representing traffic scenarios, and b) an interface for stitching together arbitrary models of traffic. Dr. Gomes will demonstrate the concept with three sample models (macroscopic, mesoscopic, and microscopic), and also show how this abstracted view of traffic dynamics enables the deployment of traffic simulators in high performance computing (HPC) environments.