具有自适应能力的仿生硬件是容错领域一个新兴的研究方向.同类细胞替换、成体干细胞分化和异类细胞转化等生物机制是人体血液组织健壮性的重要来源.受这些生物机制的启发,提出了一种名为电子组织的自适应可重构多细胞阵列结构.该结构采用了基于标记与识别的数据处理方式,解除了传统多细胞阵列结构中操作与细胞单元间的严格绑定的数据处理方式,使得电子组织具备了更为灵活的细胞单元替换能力,并在此基础上实现了同类细胞替换、成体干细胞分化和异类细胞转化3种仿生机制.这3种机制使电子组织具备了层次化的自我修复能力;成体干细胞分化和异类细胞转化机制又赋予了电子组织自我进化的能力.在FPGA上实现了原型系统,通过故障注入实验验证了原型系统的自我修复能力和自我进化能力;并通过与电子DNA结构的比较,对电子组织的自适应能力进行了分析与讨论.
In the field of fault tolerance, adaptive bio-inspired hardware is springing up in recent years. The robustness of human blood system is derived from substitution among homogeneous cells mechanism, differentiation of adult stem cells mechanism, and conversion between heterogeneous cells mechanism. Illumined by the mechanisms mentioned above, this paper presents a bio-inspired adaptive reconfigurable hardware architecture named electronic tissue (eTissue). Different from existing multicellular array, eTissue recognizes and processes data based on data tag, which loosely couples operations and processing elements, consequently equips eTissue with flexible cell replacement capability. We implement substitution among homogeneous cells, differentiation of adult stem cells, and conversion between heterogeneous cells based on this flexible cell replacement capability. These mechanisms compose the hierarchical self-healing of eTissue, and the self-evolution of eTissue is derived from differentiation of adult stem cells and conversion between heterogeneous cells. We implement the eTissue prototype system in FPGA, and conduct fault-injection experiments to attest to its self-healing and self-evolution capability. Finally, we analyze and discuss the robustness of eTissue.