Localizing, modeling, and drifting an autonomous RACECAR

Abstract

In this thesis, I discuss several topics which are core to the development of car-like autonomous systems, using the small scale autonomous RACECAR platform as a testbed. The ultimate goal of this work was to create an autonomously drifting car, which has been achieved with nonlinear trajectory optimization, and time-varying LQR trajectory tracking. Topics include a discussion of the RACECAR platform, dynamics modeling of car-like systems, robot localization, autonomous RACECAR control, and the design of a custom RACECAR simulator. In developing a high performance particle filter localization algorithm, I designed and implemented a novel algorithm for high performance two-dimensional ray casting in occupancy grid maps, called the Compressed Directional Distance Transform (CDDT). Here, I will discuss the theory and characteristics of the CDDT algorithm, which has been implemented in RangeLibc, my open-source software library.