Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.

  

Supervised Learning of the Safe set in Autonomous Driving

Author(s):

S. Maassen
Conference/Journal:

Semester Thesis, SS17
Abstract:

Guaranteeing safety in autonomous driving applications is a challenging task, due to the complexity of the models and the road constraints. Recent work showed that for autonomous racing where the road/track constraints are known a-priori the problem can be tackled by using a simplified model and viability theory. For this case the viability kernel describes all safe states. However, if the road constraints are not known a-priori this method cannot be used due to the long computation times of the viability kernel algorithm. In this thesis we study the use of supervised learning to predict the viability kernel for an arbitrary curve. Therefore, a support vector machine (SVM) is trained with a library of precomputed turns of different radii and opening angles. We demonstrate this approach by considering a library of curves motivated by the ORCA project and show that it is indeed possible to learn a general curve using a SVM. Finally, it is shown that the trained SVM classifier is able to predict the viability kernel of yet unseen curves.

Supervisors: Alexander Liniger, John Lygeros

Year:

2017
Type of Publication:

(13)Semester/Bachelor Thesis
Supervisor:



File Download:

Request a copy of this publication.
(Uses JavaScript)
% Autogenerated BibTeX entry
@PhdThesis { Xxx:2017:IFA_5651
}
Permanent link