Study On Error Detection Of Software In Space-Robot

In response to the harsh environment of outer space, space hardware and software systems must have anti-radiation characteristics of the fault-tolerant. Fault-tolerant can be achieved as the basis for other characteristics of the high dependability. In recent years, the commercial shelf (COTS) devices are increasingly widely used in modern aerospace systems, together with the size, weight and other physical conditions and cost factors, which put hardware implemented fault-tolerant hardware in difficulties. Not dependent on hardware, software implemented fault-tolerant technology is achieved as a result of high universality, low-cost and close relationship with natural COTS advantage. So it becomes a research hotspot.Space robot is universal and representative for study on error detection of high dependable software. It is a typical example of modern space systems. It can fly or navigate in the outer space. It is typical of modern aerospace systems and has many modern characteristics, such as large-scale use of COTS radiation devices in hardware and with characteristics of component-based, distributed, and modern multi-tasking operating system equipmented in software. Therefore, it is representative for study on falult telerance of high dependable software in space-robot. Traditional techniques can't meet these new features, so new techniques should be studied.Thinking about the characteristics of space robot high-dependable software systems, this paper examined the software implemented error detection techniques to achieve the basic theoretical model and evaluation system. Software fault-tolerance problem of space robot are studied in the two levels of both process and system-level respectively. And finally a unified model is used to coordinate the two-level, and a prototype system is developed for evaluation.In this paper, error detection and correction instructions are added during the compile phase for the errors in the process. In order to break the restrictions of the register allocation, the concept of analogize registers is raised. The core is using the unbanlance of the registers in the compiled code. Memory or temporarily unused registers are taken as register for added error detection or correction instructions. SEUs caused errors are divided into control flow errors and data errors. Control flow error detection algorithms and data error detection algorithms are designed based on analogize register allocation algorithms, and these two are combinatorial optimization to expand the coverage of error detection, improve the detection results. The experiments show that the method is superior to other methods available.There are two steps in the system-level error detection. First, study the framework of the model, and then the realization of different algorithms to achieve the idea. CDEDM is raised in this paper, which is suitable for distributed, component-based technologies, as well as complex multi-process software model. This is a hierarchical model, breaking the restriction of the previous studies in software fault-tolerant which can't be used for the complex software systems. Moreover, the model also makes full use of idle system resources to build the redundant system. It can effectively improve system performance, and it has good adaptability and scalability. In this paper, it is raised that the CDEDM model-based IO error detection and micro-check point error detection algorithms. IO error detection algorithm is non-invasive technology, it can check the correctness and the order of the content, while the detection based on micro-check points are more flexible. The location, density and other details of the micro-checkpoint can be set. They are both asynchronous detection. The experiments show that the two methods can be combined well.Finally, a hybrid model, MSEDS, and a prototype system are established to verify the research results of process level and system-level on the backgroud of space robot. Generally speaking, MSEDS unites the process level and system-level research results into the same system to relize full error coverage of software. Fault injection system can help to verify the results. The faults are injected into the system under the conditions of simulation of the actual space environment. The prototype system is in line with expectations. The average fault detection rate is about 93.5 percent, and the average rate for the normal operation is about 95.5 percent, while the average error correction rate is about 95.1 percent. All this costs about 21 time overhead and about 132.1 percent space overhead.