Synthesis of Reliable and Cost-Effective Cyber-Physical System Architectures
Nikunj Bajaj
EECS Department, University of California, Berkeley
Technical Report No. UCB/EECS-2015-41
May 1, 2015
http://www2.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-41.pdf
Cyber physical system (CPS) are interconnection of computation, networking and physical elements. Modern CPS are distributed, networked and safety-critical sys- tems and architectural design of such systems with fault tolerance and performance constraints is a challenging task. In this thesis, we address the problem of synthe- sizing safety-critical cyber-physical system architectures to minimize a cost func- tion while guaranteeing the desired reliability. We cast it as an optimization prob- lem with the component cost as the objective and the performance and reliability requirements as the constraints. The challenge is to generate symbolic reliability constraints, which is exponential in the size of the system due to exhaustive enu- meration of failure cases on all possible system configuration. We propose two al- gorithms to overcome this problem that we refer to as Integer-Linear Programming Modulo Reliability (ILP-MR) and Integer-Linear Programming with Approximate Reliability (ILP-AR). ILP-MR solves an easier optimization problem with perfor- mance constraints and iteratively introduces redundancy in the system with a back- ground reliability analysis routine. Conversely, ILP-AR solves the problem in one iteration by symbolically representing the reliability constraints computed using an in house developed approximate algebra. We compare the two approaches and demonstrate their effectiveness on the design of aircraft electrical power system architectures.
Advisors: Alberto L. Sangiovanni-Vincentelli
BibTeX citation:
@mastersthesis{Bajaj:EECS-2015-41, Author= {Bajaj, Nikunj}, Title= {Synthesis of Reliable and Cost-Effective Cyber-Physical System Architectures}, School= {EECS Department, University of California, Berkeley}, Year= {2015}, Month= {May}, Url= {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-41.html}, Number= {UCB/EECS-2015-41}, Abstract= {Cyber physical system (CPS) are interconnection of computation, networking and physical elements. Modern CPS are distributed, networked and safety-critical sys- tems and architectural design of such systems with fault tolerance and performance constraints is a challenging task. In this thesis, we address the problem of synthe- sizing safety-critical cyber-physical system architectures to minimize a cost func- tion while guaranteeing the desired reliability. We cast it as an optimization prob- lem with the component cost as the objective and the performance and reliability requirements as the constraints. The challenge is to generate symbolic reliability constraints, which is exponential in the size of the system due to exhaustive enu- meration of failure cases on all possible system configuration. We propose two al- gorithms to overcome this problem that we refer to as Integer-Linear Programming Modulo Reliability (ILP-MR) and Integer-Linear Programming with Approximate Reliability (ILP-AR). ILP-MR solves an easier optimization problem with perfor- mance constraints and iteratively introduces redundancy in the system with a back- ground reliability analysis routine. Conversely, ILP-AR solves the problem in one iteration by symbolically representing the reliability constraints computed using an in house developed approximate algebra. We compare the two approaches and demonstrate their effectiveness on the design of aircraft electrical power system architectures.}, }
EndNote citation:
%0 Thesis %A Bajaj, Nikunj %T Synthesis of Reliable and Cost-Effective Cyber-Physical System Architectures %I EECS Department, University of California, Berkeley %D 2015 %8 May 1 %@ UCB/EECS-2015-41 %U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-41.html %F Bajaj:EECS-2015-41