BMIRL

BMIRL = Bayesian Multitask Inverse Reinforcement Learning This project is part of my Masters' thesis at UGA and partly funded by a major car manufacturer for research on autonomous driven cars. The contribution here is the application of an algorithm which is derived from Algorithm 1 of https://arxiv.org/pdf/1106.3655.pdf on the NGSIM dataset (https://www.fhwa.dot.gov/publications/research/operations/06137/) More details can be found in the pdf here.The complete thesis titled 'Learning driver preferences for highway car merging' when published in UGA's thesis repository will be available for download at http://dbs.galib.uga.edu/cgi-bin/getd.cgi?userid=galileo&serverno=16&instcode=publ