Researches On Multi-objective Evolution Of Digital Circuit

Evolvable hardware (EHW), also named bio-inspired hardware, is a kind of hardware circuit by changing its structure to adapt the environment outside automatically and dynamically with the capabilities of self-organization, self-adaptation and self-repair. It has extensive application prospect in circuit design, automated control, fault-tolerant system, pattern recognition and artificial intelligence, robot, space and deep-sea exploration.In this thesis, aiming at the scalability issue, a new method with multi-objective optimization is proposed to evolve larger digital circuit and redundant system with different structures. As the result, digital system for fault tolerance is designed successfully with larger scale, more excellent structure and better capability of fault tolerance. The main contents of this paper are as follows:(1) The evolutionary design is based on disjunction and replacement of truth table which uses multistage decomposition to separate variables and replaces truth table with sub-truth table in order to decompose the complex circuit into simple subsystems which can be constructed larger digital circuit. The results of experiments show that this method can evolve circuits successfully in larger scale.(2) The multi-objective evolutionary design method of digital circuit is researched. Considering function, resources and time delay of the circuit, the strategy of seeking excellent compounding of input variables is adopted and multi-objective evolutionary algorithm is combined with niche technology and keeping the best individual in the archives. With the optimizing method, the optimization design of digital circuit is implemented. Experimental results show that digital circuits can be designed with less resources and time delay effectively.(3) The multi-objective evolutionary design of redundant system with different structures is researched. As the deficiency of methods existed, a reasonable strategy based on topological structures is researched. Using multi-objective optimization algorithm, redundant system with different structures is designed with better fault-tolerant performance.The work presented in this paper has been funded by National Natural Science Foundation of China (60871009) and Aeronautical Science Foundation of China (2009ZD52045).