Fault-tolerant Design Of COTS-based Multi-processor Arbitration For On-board Computer

Satellites play an important role in various fields of modern society.Because they are disturbed by space particles and radiation in the complex space environment,radiation-hardened devices and reliable system design methods are generally used to strengthen satellites.With the rapid development of satellite technology,small size,lightweight,and diversified functions have become a trend,resulting in higher requirements for satellite design cycles and functional integration.Commercial-OffThe-Shelf(COTS)devices have gradually been applied in satellite systems due to their low cost and high performance.However,the disadvantages of COTS devices are poor anti-radiation ability and easy to be affected by space radiation and particles,which can cause errors.To enhance the reliability of on-board computers that employ COTS components,this paper proposes a design of multi-processor arbitration fault-tolerant systems for on-board computers based on COTS devices.Firstly,a Markov model is used to analyze and compare different system fault tolerance schemes.The results show that the hardware fault tolerance scheme using triple modular redundancy has high reliability.The parameters affecting system reliability are then analyzed,and it is concluded that improving the reliability of the voting unit and the repair rate of the system can help improve the overall system’s reliability without enhancing the single COTS device’s reliability.Secondly,to address the issue that the key voting unit of the current triple modular redundancy system cannot detect its own errors and has limited judgment ability regarding the health status of a single computer,a strengthened key voting unit scheme,utilizing a totally self-checking design,is proposed.A combination of a multi-channel dual-rail code checker and a three-in-six code checker is employed to judge the consistency of the voting data,while a similar structure is added to distinguish the cause of the fault.During the design process,a problem arises wherein the normal situation of the data is consistent,leading to the inability to input the complete test set.To address this issue,a method is introduced in which the randomness of the data itself disrupts the input of the checker,improving the voting device’s design with totally self-checking.Finally,this paper presents a universal COTS device multiprocessor software and hardware test platform,combining the discussed hardware scheme of triple modular redundancy and the reinforcement design of the voting unit,which realizes the arbitration of single computer data and the maintenance mechanism of system health.The test results show that the proposed multi-processor arbitration fault-tolerant design can improve the on-board computer system’s reliability using COTS components,and the general design method can provide ideas for accelerating satellite development.