Multi-objective optimization of integrated circuits and systems

Prof. E. Tlelo Cuautle
Mercoledì, 15 maggio09:00 – 13:00

Designing integrated circuits using modern nano-metric technologies is challenging because a very small variation in their design parameters may deviate the desired operation and then degrade the performance of the circuit. Another issue is that integrated circuits and systems can have many target specifications that may be in conflict and then the design exploration can lead to a huge number of values for the search spaces and target specifications. This justifies the application of metaheuristics like evolutionary algorithms for multi-objective optimization. In this manner, this seminar will show how to perform multi-objective optimization of systems from their mathematical description, down to their design using MOS transistors. Other issues that can be considered in robust designs are also reviewed like process, voltage and temperature (PVT) variations, and Montecarlo simulation. It will be reviewed the basics of Evolutionary algorithms, which store a “knowledge” about the solving problem in a “population” (a set) of “individuals” (vectors of solutions). The individuals are selected, crossed, and mutated to generate a new population. The best solutions among the father’s and children’s populations will survive to the next generation (next iteration of the algorithm), which is also a metaphor of Darwin’s ideas about survival of the fittest. In addition, the seminar will review Swarm Optimization algorithms to enhance integrated circuits and systems design.