Research On Key Technologies Of Bio-Inspired Self-Organized Computing Array Architecture

High dependability design is the key issue of modern spacecraft control system. Fault-tolerant technolory is the main aspect of dependability design.There are intrinsic limitations for traditional fault-tolerant methods such as TMR(Triple Modular Redundancy).It occupies considerable resources, while its fault-tolerant capability is limited. The powerful self-adaptive ability of biological organisms brings inspiration to designing fault-tolerant method. In recent years, researchers have shown interest in bio-inspired fault-tolerant architecture. Based on the analysis of traditional fault-tolerant methods and existing bio-inspired fault-tolerant architecture, a novel bio-inspired self-organized computing array architecture is proposed in this paper.Firstly, we have studied the label-based bio-inspired computing array. By imitating four biological principles, namely, match-based recognition in protein sorting, substitution among homogeneous cells, differentiation of stem cells, and conversion between heterogeneous cells, we design the architecture of bio-inspired computing array, which supports hierarchical self-healing. We implemented the computing arrays which are specific to Median Filter algorithm and Sobel operator respectively by programming with MPI. Fault-injection experiments prove the validity of this computing array and the feasibility of its fault-tolerant approach.To solve the cell ranking problem in the bio-inspired computing array where there is no label, we have presented a distributed ranking algorithm based on the pattern formation theory and morphogen model in developmental biology. Experiments show that our method has better robustness than existing methods.We have also studied the multifunctional differentiation of computing array through two models: gene activation and lateral activation. A set of gene activators and a common gene repressor act in both single and multiple gene activation. We also find lateral activation has special size-regulation capability.At last, we have implemented the morphogen model of computing array with analog circuit, and verified it with MultiSim10.