A Dynamical Evolution Approach to High Placement Performance and Low Fault-Tolerant Cost for Bio-Inspired Self-Repairing Hardware

Bio-inspired self-repairing hardware is a reconfigurable system characterized by high fault-tolerant ability. To further increase the placement performance and reduce the fault-tolerant cost, a dynamical evolution approach to multi-objective dynamic placement is developed. A primary challenge for our study is computational complexity. Based on group theory, the computing task was divided on the dimensions of time and space to increase the utilization of online computing resources. By introducing the multi-step placement transformation, a discrete dynamical system model was built and the multi-objective dynamic placement was converted into a stability problem of constrained systems. In order to satisfy the dynamical requirements, an artificial particle system model was built and its evolution laws were verified by dynamic analysis. By simulating the evolution laws, a dynamical evolution algorithm was proposed, which could avoid large-scale iterative calculation used in the conventional optimization search algorithms. Experiments have shown that the dynamical evolution method could offer high placement performance with low fault-tolerant cost. Besides, it also has the advantage of high design flexibility since there is no initial placement constraint.