Fault-tolerant on-board evolutionary platform for adaptive allocation of hardware and software tasks

Abstract

Fails in aerospace missions, caused by broken communication and unexpected situation, have generated a pressing need for intelligent adaptation, repairing and upgrading capabilities in on-board maintenance (OBM) applications. In this paper, we present a fault-tolerant on-board evolutionary platform for aerospace applications which utilizes FPGA technology. To avoid complex evolution process of circuit and function, we take tasks as our basic chromosome and implement relocation of hardware and software tasks in spaceborne computing systems. Tasks can run in software space or be put into hardware task slots according to energy-efficient or real-time requirements. The allocation of hardware and software tasks will adapt to the requirement of current condition to meet real-time, energy-efficient or environment requirements. The process is transparent to users so that there is no need to explicitly specify reconfiguration commands. Sleeping tasks, redundant FPGAs and version switching control are combined together to achieve fault-tolerance in the on-board evolutionary platform. Design theories and strategies of FOEP are described in detail. System performance is also evaluated using simulated experiments.