MSc Project Proposal:
Parallelizing controller synthesis software using GPU’s
Project description:
There has been a recent boom on the “correct-by-design” synthesis of control
software. Many techniques have been proposed in recent years based on the idea
of constructing symbolic abstractions of control systems and employing
automated synthesis approaches on those abstractions. However, the majority of
these approaches suffer from big scalability problems. On the other hand, the
main bottleneck of these techniques tends to be the highly parallelizable task of
constructing a symbolic abstraction of the system. In this project we would like
to explore these parallelization possibilities by implementing symbolic
abstraction algorithms on a GPU.
Prerequisites:
Knowledge in the areas of non-linear control and hybrid systems is a must. Good
programming skills in Matlab and C/C++. If possible, some familiarity with GPU
programming is desired.
Some knowledge on validation/verification of systems is a plus.
References:
[1]
“Specification-Guided Controller Synthesis for Linear Systems and Safe
Linear-Time Temporal Logic” M. Rungger, M. Mazo Jr. and P. Tabuada. To
appear Hybrid Systems, Computation and Control, 2013.
[2]
“Pessoa: A tool for embedded controller synthesis” M. Mazo Jr., A.
Davitian and P. Tabuada. Computer Aided Verification, CAV2010.
Contact: Dr. Manuel Mazo Jr. <m.mazo@tudelft.nl>