Friday Seminar - Tasos Kouvelas on Adaptive Fine-tuning for Large-scale Nonlinear Traffic Control Systems

Traffic Light Tree

This week’s Friday TRANSOC Seminar has Tasos Kouvelas, Ph.D., Postdoctoral Researcher, University of California, Berkeley, presenting  “Adaptive Fine-tuning for Large-scale Nonlinear Traffic Control Systems.”

This talk introduces and analyzes a new learning/adaptive algorithm that enables automatic fine-tuning of Large-scale Nonlinear Traffic Control Systems (LNTCS), so as to reach the maximum performance that is achievable with the utilized control strategy. LNTCS have many applications in transportation, as with urban signal control or ramp metering, and yet their efficient design and deployment remains elusive due to the involved complexity and nonlinearities. Often, the deployment of a new algorithm (or the updating of an existing one) requires extensive fine-tuning before it reaches its best achievable performance. Typically, this fine-tuning procedure is conducted manually, via trial-and-error, relying on expertise and human judgment and without the use of a systematic approach. The proposed Adaptive Fine Tuning (AFT) algorithm is aiming at replacing the conventional manual optimization practice with a fully automated online procedure. The talk provides a detailed analysis of the algorithm as well as a step-by-step application description. Finally, application results of the algorithm to real-time fine-tuning problems of general LNTCS are presented using the commercial micro-simulation tool AIMSUN.

The seminar will take place today 4:00 PM in 406 Davis Hall. Please join us for a TRANSOC-sponsored Cookie Hour in the ITS Library at 3:30 PM.