*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Ph.D. Thesis Proposal
by
Kazuhide Okamoto
Advisor: Dr. Panagiotis Tsiotras
“TOWARDS HUMAN-AUTONOMY COLLABORATION:
A MACHINE-LEARNING AND STOCHASTIC OPTIMAL CONTROL-BASED APPROACH”
3:30 p.m., Wednesday, April 18
Montgomery Knight Building Room 325
Abstract:
Autonomous systems such as self-driving vehicles can be regarded as special types of robots that have to share the environment with humans. Traditional autonomous systems, which do not necessarily share the environment, have played an important role in a large range of engineering applications and have succeeded in reducing human workload, e.g., vehicle manufacturing, operation, and maintenance. However, it is still challenging for autonomous systems to share the environment and collaborate with humans. The aim of this work is to contribute towards the goal of an “intelligent” system that works as the co-pilot of the vehicle and collaborates with the human driver. To this end, we follow the following two steps: 1) understand the intentions of the driver and other vehicles in traffic, and 2) based on this understanding, execute proper actions. In this thesis proposal, we first summarize our previous work on the first task, i.e., using machine-learning methods to understand human intentions in the immediate future. Then, we introduce a newly developed stochastic optimal control method, namely, the chance-constrained optimal covariance steering. This new method steers the mean and the covariance of a random variable of a stochastic system with a guarantee that the probability of violating the state constraints is below the pre-specified threshold. After introducing the previous work, we propose a new (semi-)autonomous vehicle control system that uses machine-learning methods to understand human intentions and executes proper actions based on the optimal covariance steering method in the receding-horizon-control fashion. As uncertainties always exist in the environment, the newly-developed control system is expected to be able to more properly support human drivers than deterministic control-based approaches.
Committee Members:
