Reliability Analysis Method and Fault Tolerance Device for I / O Interface of Intelligent Mobile System

Abstract

The present invention provides a method for analyzing the reliability of an I / O interface of an intelligent mobile system and a fault-tolerant device. The method includes: establishing a reliability model RMIO; using CTMC to describe the RMIO model, establishing a mapping relationship, and constructing a CTMC IO model; according to CTMC IO evaluate the reliability of each functional module of the I / O interface of the intelligent mobile system, and calculate the instantaneous availability of each functional module; conduct a reliability assessment on the overall interface; for the functional modules with relatively low reliability, use an I / O interface error detection scheme based on task-level redundancy to detect possible faults and errors of the functional modules; use a repair method based on reinforcement learning to repair the interface hardware system IOHS in a failed state. The present invention describes the structure and composition of the I / O interface of the intelligent mobile system and its impact on the overall reliability, strictly analyzes and evaluates the reliability of the interface, and preferentially tolerates faults in the modules with relatively low reliability to minimize the delay of fault repair.

Description

Technical Field

[0001] The present invention belongs to the technical fields of reliability analysis and fault tolerance, and particularly relates to a method for analyzing the reliability of an I / O interface of an intelligent mobile system and a fault tolerance device. Background Art

[0002] The I / O interface of an intelligent mobile system, as an important part of the intelligent mobile system, is a software and hardware interface for the intelligent mobile host and peripheral devices to interact. Generally speaking, the information exchange between the host and the outside world is carried out through input / output devices (Input / Output device), and general input / output devices are mechanical or electromechanical combinations. Compared with the high-speed central processing unit (Central Processing Unit, CPU), their information processing speed is much slower. In addition, the signal forms and data formats of different peripheral devices are also different. Therefore, peripheral devices cannot be directly connected to the CPU, but need to complete the speed matching, signal conversion between them through corresponding circuits and software protocols, and implement certain control functions. This kind of buffer circuit and software protocol is called the I / O interface (Input / Output interface) of the intelligent mobile system.

[0003] The I / O interface of the intelligent mobile system specifically includes the following functions: ① data buffering, that is, realizing speed matching; ② data format conversion; ③ level matching and time coordination; ④ exchanging control / status information. The I / O interface of the intelligent mobile system can generally be divided into: (1) parallel interface and serial interface; (2) synchronous interface and asynchronous interface; (3) direct program control, program interrupt and direct memory access (DMA) interface. The I / O interface of the intelligent mobile system consists of a register group, a control logic circuit, signal connection lines, data address lines, and control status signal lines between the host and the interface and between the interface and the I / O device. The typical structure of the I / O interface of the intelligent mobile system is as shown in Figure 2 (Typical Structure Diagram of the I / O Interface of the Intelligent Mobile System).

[0004] Traditional hardware reliability technologies include reliability block diagrams, fault trees, event trees, etc. However, with the development of intelligent mobile systems, the hardware structure of the I / O interface of intelligent mobile systems has become increasingly complex, and traditional methods are now difficult to describe its dynamic characteristics and the correlation between its reliability and time. Since the Markov model can reflect the correlation between the target object and time through state transitions and can well study dynamic characteristics, the Markov model is introduced.

[0005] Fault tolerance techniques generally include hardware-implemented fault tolerance (HIFT) and software-implemented fault tolerance (SIFT). HIFT techniques include ECC (Error Correcting Code), triple module redundancy (TMR), and many techniques proposed in the literature to improve fault tolerance capabilities that require redundant peripheral hardware devices, which may incur high hardware and reconfiguration costs. Software redundancy techniques do not require additional hardware devices, have lower implementation costs, and are more flexible, usually requiring little modification to the existing hardware structure. However, in some cases, pure software-implemented fault tolerance solutions cannot provide sufficient fault coverage. Information redundancy techniques generate supplementary information based on the original information and add the supplementary information to the original information. Information redundancy techniques include checksums, parity bits, ECC codes, etc. Time redundancy techniques are based on the additional execution time of the application and execute multiple times for fault tolerance. Fault tolerance methods based on time redundancy can be implemented at different levels: such as instruction level, program level, operating system level, and task level. Summary of the Invention

[0006] The purpose of the present invention is to provide a method for analyzing the reliability of an I / O interface of an intelligent mobile system and a fault tolerance device, which is used to describe the reliability of the I / O interface hardware of the intelligent mobile system in the intelligent mobile system, has good scalability, and can perform intelligent fault tolerance repair according to the analysis result of the I / O interface reliability.

[0007] The technical solution for achieving the purpose of the present invention is as follows:

[0008] In a first aspect, the present invention proposes a method for analyzing the reliability of an I / O interface of an intelligent mobile system, and the method includes the following steps:

[0009] Extract the reliability-related information in the design of the I / O interface of the intelligent mobile system, perform hierarchical modeling, and establish a reliability model RMIO for the I / O interface of the intelligent mobile system from bottom to top in units of the I / O interface function modules of the intelligent mobile system; by extracting the hardware structure information in the I / O interface hardware system (IOHS) of the intelligent mobile system, divide the IOHS into different I / O interface function modules (IOFM) of the intelligent mobile system, and the IOFM includes a receive data register (RDR), a transmit data register (SDR), a control register (CR), and a status register (SR).

[0010] Use a continuous-time Markov chain (CTMC) to describe the RMIO model, establish a mapping relationship, and construct a CTMC IOModel, establish a CTMC for each I / O interface function module (IOFM) of the intelligent mobile system IO Model;

[0011] Adopt the Markov chain Monte Carlo algorithm (MCMC) to randomly sample the failure rate of the I / O interface function module of the intelligent mobile system, and evaluate the reliability of each function module of the I / O interface of the intelligent mobile system according to the CTMC IO and calculate the instantaneous availability of each function module of the I / O interface of the intelligent mobile system based on this;

[0012] Describe the evolution process of the state of the I / O interface hardware system (IOHS) of the intelligent mobile system over time as a Markov chain (CTMC IO , and complete the reliability evaluation of the overall IOHS based on this Markov chain; the normal or failed states of all function modules constitute the state space S IO of the overall IOHS IO , and the state transition probability matrix A IO = [a ij is calculated through the relationship between states and the reliability evaluation results of function modules;

[0013] According to the overall reliability evaluation results of the IOHS, use the I / O interface fault tolerance scheme based on task redundancy to reinforce the IOFM with reliability lower than the set threshold, generate a manager task to create twin I / O tasks, and compare the execution results of the twin I / O tasks to complete fault detection;

[0014] On the basis of the overall reliability evaluation of the I / O interface hardware system (IOHS) of the intelligent mobile system, use the I / O interface repair method based on reinforcement learning to repair the I / O interface of the intelligent mobile system in the failed state to restore it to the normal state in the shortest time; first, take the failed state as the initial state in the state space, formulate a deterministic policy and a reward function, and finally use the ε-greedy algorithm to determine the optimal policy, and select the IOFM for repair according to the optimal policy.

[0015] Furthermore, establish the reliability model of the I / O interface function module of the intelligent mobile system, which is represented as the following triple:

[0016] RMIO IOFM = (FR IO , SS IO , STR IO )

[0017] In the formula, RMIO IOFM is the I / O interface function reliability model of the intelligent mobile system, FR IO represents the failure rate of the I / O interface function module (IOFM) of the intelligent mobile system; SSIO Represents the state space of the IOFM; STR IO Represents the state transition relationship, where STR IO Contains the source state sStateIO and the target state tStateIO; The failure rates of the receive data register RDR, transmit data register SDR, control register CR, and status register SR of the IOFM are λ RDR , λ SDR , λ CR , λ SR , respectively. Let the importance weights of the IOFM in different intelligent mobile systems be ω RDR , ω SDR , ω CR , ω SR , respectively. The calculation formula for the failure rate λ IOFM of the IOFM is:

[0018] λ IOFM = 1 - Π i∈(RDR,SDR,CR,SR) (1 - ω i λ i ).

[0019] Furthermore, the continuous-time Markov chain CTMC IO is a five-tuple:

[0020] CTMC 10 = (S IO , S in , A IO , T IO , t)

[0021] Where, S IO represents the state space of CTMC IO , S in ∈ S IO represents the initial state of CTMC IO , A IO = [a ij represents the state transition probability matrix, a ij represents the probability of transitioning from state S i ∈ S IO to state S j ∈ S IO ; is the set of state transition relationships, (S i , S j ) represents the existence of a transition from state S i to state S j ; t represents time.

[0022] Furthermore, the RMIO is described as CTMC IO , and by designing the RMIO and CTMCIO Perform the conversion according to the element mapping rule between them, and the mapping rule is shown in Table 1 below:

[0023] Table 1 RMIO Modeling and CTMC IO Mapping Rule between Elements

[0024]

[0025]

[0026] Furthermore, the failure rate λ of the IOFM IOFM The functional relationship with time is:

[0027]

[0028] The typical state transition matrix of the functional module of the I / O interface of the intelligent mobile system is:

[0029]

[0030] The meanings of the parameters therein are shown in Table 2 below:

[0031] Table 2 Parameters of the IOFM State Transition Relationship Diagram

[0032]

[0033]

[0034] Thus, the state probability equation of the functional module of the I / O interface of the intelligent mobile system can be obtained as:

[0035]

[0036] Among them, P N (t), P R (t), P DG (t), P D (t), P F (t), P E (t) respectively represent the probabilities that the IOFM is in the normal state, recovery state, degraded error-tolerant working state, fault detection state, failure state and fault state at time t, and t′ is the next moment of time t;

[0037] The reliability R IOFM The evaluation calculation formula of the (t) of the I / O interface functional module is:

[0038] R IOFM (t) = 1 - P E (t)

[0039] Among them, R IOFM(t) refers to the probability that the system is not in the failure state at time t.

[0040] Furthermore, when evaluating the reliability of the I / O interface hardware system of the intelligent mobile system, the state space of the overall I / O interface of the intelligent mobile system is composed of the states of all functional modules, that is, a state sequence s of all functional modules n-1 s n-2... s 1 s 0 constitutes a state of the overall I / O interface of the intelligent mobile system, where s j represents that the functional module IOFM j is in the normal state s j = 0 or the failure state s j = 1, 0 ≤ j < n, and n is the number of functional modules.

[0041] In the second aspect, the present invention proposes a fault-tolerant device for the I / O interface of an intelligent mobile system. The device includes:

[0042] An extraction module, configured to monitor the states of each functional module of the I / O interface of the intelligent mobile system, extract the hardware structure information in the IOHS of the intelligent mobile system, and obtain the state information of the functional modules; the state information includes the data saved in the receive data register, the data saved in the transmit data register, the control bit information of the control register, the status bit information in the status register, and the importance information;

[0043] A model construction and conversion module, configured to construct a reliability model RMIO of the functional modules of the I / O interface of the intelligent mobile system IOFM , convert the RMIO IOFM model into a CTMC IO , and perform model element mapping through a pre-designed element mapping rule to generate a continuous-time Markov model;

[0044] A reliability evaluation module, configured to evaluate the overall reliability of the IOHS according to the generated CTMC IO , and calculate the states of the overall state space of the IOHS using the relationships between the states and the reliability evaluation interfaces of the functional modules;

[0045] A fault-tolerant module, configured to repair the I / O interface of the intelligent mobile system in the failure state according to the overall reliability evaluation result of the IOHS, using an I / O interface error detection scheme based on task redundancy and an I / O interface repair method based on reinforcement learning.

[0046] Further, the fault tolerance module creates a manager task using an I / O interface error detection method based on task redundancy and generates two twin versions of the I / O task by the manager task; the manager task is responsible for input preparation, scheduling, output comparison, and error recovery operations of the twin tasks; first, the first version and the second version of the twin tasks are run. If the execution results of these two twin tasks are different, the third version of the twin task will be run. Once the manager task detects that the outputs of the first two versions of the twin tasks do not match, it will save the task results of the first version and the second version, wait for the third version task to finish execution, judge the correct execution result of the task, and finally restore the state of the faulty task stack to the initial state to prepare for the operation of the next task.

[0047] Further, the I / O interface repair method based on reinforcement learning used by the fault tolerance module is constructed by the following method:

[0048] In the first step, a unique encoding is created for all state distributions of the IOHS. This encoding can not only be used to distinguish all states of the IOHS but also be associated with the reliability of the IOHS; for each state of the IOHS, there is a unique encoding corresponding to it, and this unique encoding is composed of the states of all IOFMs in the IOHS. The state sequence of all IOFMs is used as the state encoding of the IOHS.

[0049] In the second step, the state received by the IOHS from the external environment at time t is S t ∈S, and the operation that the IOHS can select from the set of repair operations allowed by the current state at the current moment is denoted as M t ∈M(S t ). After taking the repair operation, the state of the IOHS at the next moment is S t+1 ∈S + , and the reliability repair value brought by this repair operation is where all states are denoted as S + , and non-terminal states are denoted as S;

[0050] In the third step, the characteristic equation obtained according to the Markov decision process is as follows:

[0051] p(s′,r|s,m)=p r [S t+1 =s′,R t+1 =r|S 0 ,M 0 ,S 1 ,M 1 ,...,S t ,M t

[0052] ​Among them, p(s′, r|s, m) is the probability of transitioning to state s′ when considering the reliability repair value r in state s and taking the repair operation m ∈ M(S); this probability is determined by the state of the IOHS and the set of repair operations S 0 , M 0 , S 1 , M 1 ,..., S t , M t ∈(S, M(S)), which represents the conditional probability that the state at time t + 1 is s and the reliability repair value at time t + 1 is r, where the subscript r represents the reliability repair value;

[0053] The decision-making process DP is defined as the following five-tuple

[0054] DP = (S, M, P, R, γ)

[0055] where S represents the state space; M represents the set of repair operations, which is finite; P represents the state transition probability from the initial state to the target state; R represents the IOHS reliability repair value obtained after taking the repair operation; γ ∈ [0, 1] represents the attenuation rate, which quantifies the difference in reliability repair values;

[0056] Fourthly, the value function of the repair model is as follows

[0057]

[0058] where R(s IO (k), m IO (k)) is the reward function, that is, for the state s IO (k) at time k, taking the repair operation m IO (k) brings the reward, and γ k is the cumulative reliability repair value attenuation rate from the initial time to time k; multiplying the discount factor and the reward function and summing over the repair steps is the final objective function. Select the repair step that can produce the maximum objective function. By making the best action at each state, the final repair step is obtained;

[0059] Fifthly, taking all the states of the IOHS as the state transition set, set the failure state of the IOHS as the initial state; for the IOFM failure caused by instantaneous faults, repair it through fault-tolerant methods such as refreshing reconfiguration and waiting for rewriting. For the IOFM failure caused by permanent faults, use methods such as hardware-implemented multi-mode redundancy and component replacement for repair; use the ε-greedy algorithm to calculate and take the result of the formula in the fourth step to obtain the repair action with the maximum value. Select the IOFM to be repaired from the state transition set and take the optimal action to repair it, and then repeat using this repair action until the IOHS returns to the normal state.

[0060] Compared with the prior art, the significant advantages of the present invention are as follows:

[0061] 1) The present invention adopts a hierarchical modeling form, taking the I / O interface function module of the intelligent mobile system as a unit to establish a reliability model for the I / O interface of the intelligent mobile system, which is easy to understand, simple and clear, and the model will not be bloated.

[0062] 2) It has strong data constraint ability and good scalability, can describe probabilities, and can conveniently add reliability elements.

[0063] 3) It designs an I / O interface error detection scheme based on task redundancy and an I / O interface fault tolerance method and device based on reinforcement learning, which can effectively repair the I / O interface in a failed state.

[0064] The present invention will be further described in detail below with reference to the accompanying drawings. Description of the Drawings

[0065] Figure 1 It is a flowchart of a reliability analysis method and a fault tolerance device for an I / O interface of an intelligent mobile system.

[0066] Figure 2 It is a hardware abstraction diagram of an intelligent mobile system centered on the I / O interface of the intelligent mobile system.

[0067] Figure 3 It is a state transition relationship diagram of an I / O interface hardware system IOHS of four I / O interface function modules IOFM of an intelligent mobile system.

[0068] Figure 4 It is a state transition diagram of a function module IOFM of a typical I / O interface of an intelligent mobile system.

[0069] Figure 5 It is an architecture diagram of an I / O interface fault tolerance device for an intelligent system. Detailed Embodiments

[0070] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0071] In one embodiment, the present invention proposes a reliability analysis method for an I / O interface of an intelligent mobile system, which specifically includes the following steps:

[0072] Step S101: Extract the reliability-related information in the I / O interface design of the intelligent mobile system, and perform hierarchical modeling. Taking the I / O interface function module of the intelligent mobile system as a unit, establish a reliability model RMIO (Reliability Model of I / O interface) for the I / O interface of the intelligent mobile system from bottom to top. By extracting the hardware structure information in the I / O interface hardware system (IOHS) of the intelligent mobile system, divide the IOHS into different I / O interface function modules (IOFM) of the intelligent mobile system. The IOFM includes a receive data register RDR (Receive Data Register), a send data register SDR (Send Data Register), a control register CR (Control Register), and a status register SR (State Register). The failure rates of the above function modules are represented by λ RDR 、λ SDR 、λ CR 、λ SR respectively. Since the impact of each I / O interface function module of the intelligent mobile system on the overall reliability of the I / O interface of the intelligent mobile system is different in different intelligent mobile systems, four importance weights ω RDR 、ω SDR 、ω CR 、ω SR of the I / O interface function modules of the intelligent mobile system are introduced;

[0073] Step S102: Use the continuous-time Markov chain CTMC (Continuous Time Markov Chain) to describe the RMIO model, and establish a CTMC IO model for each I / O interface function module IOFM of the intelligent mobile system. The CTMC IO describes the state space S IO 、the initial state S in ∈S IO 、the state transition relationship and the function relationship matrix A IO =[a ij ;

[0074] Step S103: Use the Markov chain Monte Carlo algorithm MCMC for random sampling, use the exponential model to describe the relationship between the I / O interface failure rate of the intelligent mobile system and time, and according to the CTMC IOEvaluate the reliability of each functional module of the I / O interface of the intelligent mobile system, and calculate the instantaneous availability of each functional module of the I / O interface of the intelligent mobile system based on this;

[0075] Step S104, describe the evolution process of the state of the I / O interface hardware system IOHS of the intelligent mobile system over time as a Markov chain CTMC IO , and complete the reliability evaluation of the overall IOHS based on this Markov chain. The normal or failed states of all functional modules constitute the state space S of the overall IOHS IO , S IO The state transition probability matrix A IO = [a ij is calculated through the relationship between states and the reliability evaluation results of functional modules in step S102;

[0076] Step S105, based on the overall reliability evaluation of the I / O interface hardware system IOHS of the intelligent mobile system, use the repair method of the I / O interface of the intelligent mobile system based on reinforcement learning to repair the I / O interface of the intelligent mobile system in the failed state.

[0077] The following is a more detailed description.

[0078] An I / O interface reliability analysis method and fault-tolerant device for an intelligent mobile system proposed by the present invention establish an RMIO model. The definition of the RMIO model is given below, and a description method for various constraints in the intelligent mobile system software is given for strict analysis and verification. Constraining the attributes of modeling elements can make the established model consistent and is conducive to improving the correctness and efficiency of software modeling.

[0079] 1. Hardware structure of the I / O interface of the intelligent mobile system

[0080] Since the I / O interface of the intelligent mobile system serves as a bridge for communication and interaction between the intelligent mobile system and peripheral devices, in order to study the SEU resistance of the I / O interface of the intelligent mobile system and accurately model the reliability of the I / O interface hardware system (I / O interface Hardware System, IOHS) of the intelligent mobile system, it is necessary to first clarify the structural model of the IOHS. The IOHS can be divided into different I / O interface function modules (I / O interface Function Module, IOFM) according to functions, and the IOFM is composed of various mechanical and electronic components. Therefore, based on electronic components and IOFM, an abstract definition and description of the structure of the IOHS can model and analyze the reliability of the IOHS from the essence. Such as Figure 2As shown, the I / O interface of the intelligent mobile system can be divided into a Receive Data Register (RDR), a Send Data Register (SDR), a Control Register (CR), and a State Register (SR) according to functions and specifications. The RDR is used to receive data sent by peripheral devices and transfer it to the processor for processing; the SDR is used to store data sent from the processor and send it to peripheral devices; the CR is used to control the level matching and timing coordination between the processor and peripheral devices; the SR is used to store the status and control information during the interaction between the processor and peripheral devices.

[0081] Definition 1. IOFM is represented as a quadruple, where each element represents the set of all corresponding elements that make up IOFM.

[0082] IOFM = (RDR, SDR, CR, SR)

[0083] Definition 2. IOHS is represented by a set, where IOHS may contain one or more CFM.

[0084] IOHS = {IOFM 1 , IOFM 2 , K, IOFM n}

[0085] It can be seen from Definition 2 that the main constituent element of IOHS is IOFM. Therefore, for the research on the reliability of IOHS, it is mainly to study the reliability of IOFM. Moreover, since IOFM is composed of underlying electronic components, studying its reliability can essentially analyze the reliability of IOHS.

[0086] 2. Reliability Model of the I / O Interface Function Module of the Intelligent Mobile System

[0087] Definition 1 gives the definition of the IOFM sub-model in the RMIO model, which details the modeling elements that make up the IOFM reliability model. The specific definition is as follows:

[0088] Definition 3. The IOFM sub-model RMIO of RMIO IOFM is represented as a triple shown in the following formula, where RMIO IOFM (Reliability Model of I / O interface IOFM ) is the functional reliability model of the I / O interface of the intelligent mobile system; FR IO (Failure Rate IO ) represents the failure rate of the I / O interface function module IOFM of the intelligent mobile system; SS IO(State Space IO ) represents the state space of the IOFM; STR IO (State TransferRelationship IO ) represents the state transfer relationship.

[0089] RMIO IOFM =(FR IO , SS IO , STR IO )

[0090] The following separately introduces the specific meanings of the three elements of RMIO IOFM and the definition methods.

[0091] (1) Failure rate FR IO

[0092] According to Definition 1, the IOFM is a set composed of the receive data register RDR, the send data register SDR, the control register CR, and the status register SR. Therefore, the reliability of the IOFM is also related to these four components, and the occurrence of a failure (Error) in any one of these four components can be regarded as a failure (Error) of the entire BFM. Therefore, the following Definition 4 can be given for the error rate of the BFM.

[0093] Definition 4. The failure rates of the RDR, SDR, CR, and SR of the IOFM are λ RDR , λ SDR , λ CR , λ SR respectively. Each type of IOFM has different impacts on the overall reliability of the I / O interface of the intelligent mobile system in different embedded systems. Therefore, the importance weights ω RDR , ω SDR , ω CR , ω SR are introduced. The values of the importance weights are determined by the specific system. Then the calculation formula for the failure rate λ IOFM of the IOFM is:

[0094]

[0095] Combined with Definition 4, the failure rate FR IO of the IOFM is expressed as follows:

[0096]

[0097] Among them, RDR_FR IO , SDR_FR IO , CR_FR IO , SR_FR IO respectively represent λ RDR, λ SDR , λ CR , λ SR , IOFM_FR IO represents the failure rate of IOFM, and there are multiple IOFMs in IOHS. Add <iofmname>For distinction.

[0098] (2) State space SS IO

[0099] The state space of IOFM includes Normal State (NS), Error State (ES), Failure State (FS), Detected State (DS), and Recovery State (RS), where NormalState is the initial state; ErrorState is a transient state and a transitional state when a fault occurs in NormalState; DetectedState is also a transient state and is the state where BFM detects a fault and further processes the combined fault. Also, for BFM with a fault tolerance mechanism, there is also a Degraded Working State (DGS). Thus, Figure 4 The typical state transition diagram of IOFM as shown, where the ellipses represent states and the connected lines with parameters represent state transition relationships and their transition probabilities. The definitions of the parameters in the figure are shown in Table 3.

[0100] Table 3 Parameters of the IOFM State Transition Relationship Diagram

[0101]

[0102]

[0103] After clarifying the state space of IOFM, its definitions are as follows:

[0104]

[0105] Among them, isInitial and isArrive respectively represent whether it is the initial state and whether it is the current state.

[0106] (3) State transition relationship STR IO

[0107] Figure 4 It not only describes the state space of IOFM but also describes the transition relationships between all states of IOFM. A state transition must declare three elements, namely the source state sState, the target state tState, and the transition probability parameter TRate. Among them, sState and tState must be included in the state space of this IOFM, and TRate comes from the transition parameters listed in the above table. The specific definition method is as follows:

[0108]

[0109] After modeling each reliability element of the IOFM, it is necessary to integrate the reliability constraints belonging to the same IOFM to facilitate hierarchical management of all reliability constraints when the IOHS is too complex and contains too many IOFMs. The local failure of the I / O interface of the intelligent mobile system can be repaired by the intelligent mobile system in software ways, such as reinstalling the driver program, hot-plugging of peripheral devices, etc. In the I / O interface system of the intelligent mobile system, when the local failure of the hardware, i.e., the IOFM failure, there is still a possibility of recovering to the normal state, and this possibility is called the repair probability of the IOFM, which can be calculated by the following formula, where μ IOFM represents the repair probability of the IOFM, F IOFM represents the number of repairs of the IOFM, Δt IOFM represents the total repair time of the IOFM, ΔT IOFM represents the total running time of the IOFM.

[0110]

[0111] Therefore, it is necessary to manage the reliability constraints belonging to the same IOFM. Use IOFMRRate to represent the repair rate of the IOFM, and the specific definition method is as follows:

[0112]

[0113] The above <iofmname>Definition method of the modes included in the IOFM.

[0114] 3. Reliability Model of I / O Interface of Intelligent Mobile System

[0115] IOFM is the basic unit of IOHS. Therefore, whether the IOFM is in a failure state directly reflects the state of IOHS. Figure 3 Describes the state transition relationship of an IOHS with four IOFMs. As Figure 3 shown, the IOHS has four IOFMs, namely the Receive Data Register (RDR), the Send Data Register (SDR), the Control Register (CR), and the State Register (SR). The circles in the figure represent a state of the IOHS. For example, the state "RDR·SDR·CR·SR" means that RDR, SDR, CR, and SR are not in a failure state in this state; "RDR·SDR·CR_" means that RDR, SDR, and CR are not in a failure state, while SR is in a failure state. λ RDR 、λ SDR 、λ CR 、λ SR respectively represent the failure probabilities of RDR, SDR, CR, and SR. Using μ RDR 、μ SDR 、μ CR 、μ SR respectively represent the repair probabilities of the four.

[0116] As Figure 3 shown, an IOHS with four IOFMs has 2 4 states. The number of IOFMs included in the IOHS directly affects the size of the state space of the IOHS. The specific relationship is that assuming the IOHS contains n BFMs, then the state space of the BAS contains 2 n states. Therefore, it is necessary to clearly declare the IOFMs included in the IOHS and their numbers. The declaration method of the IOHS is as follows:

[0117]

[0118] Among them, IOFM_num represents the number of IOFMs contained in the IOHS, and includes all IOFMs belonging to the IOHS in the declaration method of the mode inclusion.

[0119] The following specifically describes the reliability evaluation method based on the RMIO model.

[0120] The change of the state space of the RMIO model is time - related. However, the state of the RMIO model at the next moment is still only related to the current state. Therefore, the RMIO model conforms to the properties of CTMC. The definition of CTMC is given below IO as follows:

[0121] Definition 5. CTMC IO is a five - tuple as shown in the following formula:

[0122] CTMC IO =(S IO , S in , A IO , T IO , t)

[0123] where S IO represents the state space of CTMC IO ; S in ∈S IO represents the initial state of CTMC IO ; A IO =[a ij represents the state transition probability matrix, and a ij represents the probability of transitioning from state S i ∈S IO to state S j ∈S IO ; is the set of state transition relations, and (S i , S j ) represents the existence of a transition from state S i to state S j ; t represents time.

[0124] To evaluate the reliability of RMIO using CTMC IO , it is necessary to convert RMIO into CTMC IO . In order to make the conversion process equivalent and ensure that the reliability constraints remain consistent before and after the conversion, it is necessary to analyze and compare the elements contained in both and perform mapping conversions on the same elements in both. The following gives the conversion rules for the elements between the two, as shown in Table 4:

[0125] Table 4 RMIO Modeling and CTMC IO Element Mapping and Conversion Rules

[0126]

[0127]

[0128] As can be seen from the above table, all reliability constraints in RMIO are mapped to CTMC IO Among them, the reliability constraints are not omitted or modified during the conversion process, indicating that the conversion process is equivalent.

[0129] 4. Reliability Evaluation of a Single IOFM

[0130] Due to CTMC IO The final probability distribution tends to the same stable probability distribution. The stable probability distribution to which the state transition matrix of the Markov chain model converges is independent of our initial state probability distribution. We can use the state transition matrix of the Markov chain model corresponding to this stable probability distribution to quickly obtain samples that conform to the corresponding stable probability distribution. Therefore, Markov Chain Monte Carlo (MCMC) sampling is introduced. The following gives the IO concept of the stationary probability distribution of CTMC:

[0131] Definition 6. If the state transition matrix A IO of CTMC IO and the probability distribution P stable (t) satisfy the relationship shown in the following formula for all t, then P stable (t) is called the stationary probability distribution of the state transition matrix A IO .

[0132] P stable (t 1 )A IO = P stable (t 2 )A IO

[0133] By using MCMC sampling, a corresponding sample set can be found from the stationary distribution P stable (t). The process of solving P stable (t) is divided into the following steps:

[0134] Step 1, input the arbitrarily selected state transition matrix B IO of CTMC IO and the stationary distribution P stable (t), and set the state transition times threshold n 1 and the required number of samples n 2 . Among them, the state transition times threshold n 1 represents that CTMC IO can obtain the stationary distribution P 1 after n stable rounds.

[0135] Step 2, sample the initial state value S 0 from an arbitrary simple probability distribution (such as a Gaussian distribution);

[0136] Step 3, set \(i = 1\), sample from the conditional probability distribution \(B(t|t i )\) to obtain the sample \(t * \). From the uniform distribution \(u\sim uniform[0,1]\), if \(u\lt\alpha(t i ,t * )=\pi(t * )\cdot B IO (t * ,t i )\), then accept the transition \(t i \to t * \), that is, \(t i+1 = t * \), otherwise do not accept the transition, \(i=\max(i - 1,0)\);

[0137] Step 4, repeat Step 3 until \(i = n 1 +n 2 \), and the finally obtained sample set is the sample set corresponding to the final stationary distribution.

[0138] Among all the hardware module failure rate models, the exponential model has great advantages and has a high fitting degree for the true curve of the failure rate of the I / O interface module of the intelligent mobile system. Therefore, the relationship between the failure rate and time of the IOFM can be defined as follows:

[0139] Definition 7. The relationship between the failure rate and time of the IOFM is shown in the following formula, where \(\lambda IOFM (t)\) represents the failure rate of the IOFM at time \(t\), and \(\lambda IOFM \) represents the error rate at the initial time \(t = 0\), that is, the IOFM_FR IOFM declared in the FR IO of the RMIO IO .

[0140]

[0141] Let \(P IOFM (t)=(P N (t),P R (t),P DG (t),P D (t),P F (t),P E )\) represent the state probability vector of the IOFM at time \(t\), where \(P N (t)\), \(P R (t)\), \(P DG (t)\), \(P D (t)\), \(P F (t)\) and \(P E (t) represents the probabilities that the IOFM is in NormalState, RecoveryState, Degrade State, DetectedState, FailureState, and ErrorState at time t. According to the state transition equation of CTMC, the state probability equation of the IOFM can be obtained, and its calculation method is shown as follows:

[0142] P IOFM (t′) = P IOFM (t) × A IO

[0143] Among them, P IOFM (t′) = (P N (t′), P R (t′), P DG (t′), P D (t′), P F (t′), P E (t′)) represents the state probability vector of the IOFM at time t′, where t′ represents the next moment of time t, and A IO represents the state transition probability matrix of the IOFM, which can be obtained from Figure 4 as shown in the following formula.

[0144]

[0145] The state probability equation of the IOFM can be obtained as shown in the following formula:

[0146]

[0147] Solving the above formula can obtain the state probability distribution of the IOFM at time t, and the calculation method of the reliability of the IOFM at time t is shown as follows:

[0148] R IOFM (t) = 1 - P E (t)

[0149] 5. Overall Reliability Evaluation of the I / O Interface of the Intelligent Mobile System

[0150] The failure probability of the functional modules of the I / O interface of the intelligent mobile system and the RMIO IOFM in <iofmname>The IOFMRRate defined by IOFM is filled into Figure 3 Because at this time, the functional relationship between the reliability of all IOFMs and time has been calculated, that is Figure 3 the probability of each state transition in Figure 3 at any moment has been determined, and

[0151] For the I / O interface of the intelligent mobile system, only when all IOFMs are in the normal state, the entire I / O interface of the intelligent mobile system is available. Therefore, the normal state requirement of the I / O interface of the intelligent mobile system is that no IOFM is in the failure state, that is, corresponding to Figure 3 the state "RDR·SDR·CR·SR" in

[0152] Therefore, the problem of evaluating the reliability of the I / O interface of the intelligent mobile system is transformed into calculating the probability that each IOFM of the I / O interface of the intelligent mobile system is in the normal state by using the Markov chain, that is, the probability of being in the state "RDR·SDR·CR·SR".

[0153]

[0154] where P stable represents the probability distribution when the Markov chain corresponding to IOHS is in the steady state, P stable contains the probability of IOHS being in each state, and n represents the number of transitions. It should be especially noted that t in the above formula does not simply represent time, but represents the moment corresponding to the reliability of each IOFM. The transition probability of the transition matrix is based on the reliability of the IOFM, and the reliability of the IOFM is related to time.

[0155] 6. Overall Architecture of the Fault-Tolerant Device for the I / O Interface of the Intelligent Mobile System

[0156] The fault-tolerant device for the intelligent mobile I / O interface uses a task-level redundancy-based I / O interface fault-tolerant scheme and a reinforcement learning-based repair method. The architecture of this fault-tolerant device is as Figure 5 shown, and is divided into three levels as a whole. These three levels are: the core function layer, the basic software layer, and the operating system and hardware layer. The three levels are responsible for different functions respectively:

[0157] (1) Core Function Layer

[0158] The core function layer includes two subsystems: the interface detection subsystem and the fault repair subsystem.

[0159] The I / O interface reinforcement subsystem uses an I / O interface error detection scheme based on task redundancy, and is responsible for generating manager tasks and invoking manager tasks to implement the soft reinforcement function of the I / O interface based on task-level redundancy. The I / O interface reinforcement subsystem includes three main modules: generating manager tasks, comparing twin task results, and fault detection. Among them, the generating manager task mainly implants the manager task into the I / O interface driver; the twin task result comparison can use the manager task to perform task-level redundancy on the I / O task, and compare and vote on the execution interfaces of the twin version tasks, and finally output the correct I / O task result; the fault detection module determines whether a fault has occurred according to the twin task comparison result. If a fault occurs, the fault location and error information will be sent to the user in the form of a log.

[0160] The interface repair subsystem adopts an I / O interface repair method based on reinforcement learning. First, it creates a unique encoding for all state distributions of the IOHS. This encoding can not only be used to distinguish all states of the IOHS, but also must be able to be associated with the reliability of the IOHS; secondly, it defines the repair operation set and system state monitoring of the repair model, and calculates the value function of the repair model; finally, it obtains the optimal state sequence of state transition through the ε-greedy algorithm, and takes corresponding repair operations according to this sequence to repair the fault and restore the IOHS to the normal working state.

[0161] (2) Basic software layer

[0162] The basic software layer is the tool software required by the system, provides function interfaces for the upper-layer software, and realizes the integration framework for the interaction between the upper-layer software and the lower-layer software. In this layer, the mutual calls between the upper-layer function modules and the data and interfaces of the lower-layer software (such as GDB, Keil, RealEvo-IDE, etc.) are realized, and the information interaction between modules is also realized.

[0163] (3) Operating system and hardware layer

[0164] The operating system and hardware layer provides operating system support for the upper-layer software and application layer, including interface drivers, real-time operating system SylixOS, and board support package BSP. This layer abstracts hardware resources into system call functions and interfaces for modules in the basic software layer to call. The operating system and hardware layer is the bottommost and most basic layer of the entire system. The hardware includes LIN / SCI and NIC, etc. The hardware units in the hardware layer are the objects of error detection and fault recovery, and are also the basic core of the overall system.

[0165] 8. I / O interface error detection scheme based on task redundancy

[0166] The error detection scheme creates manager tasks and generates two redundant versions of I / O tasks by the manager tasks. The manager tasks are responsible for input preparation, scheduling, output comparison of the twin tasks, and error recovery operations. To improve the performance of this fault-tolerant scheme, the first version and the second version of the twin tasks are run first. If the execution results of these two twin tasks are different, the third version of the twin tasks will be run. Once the manager tasks detect that the outputs of the first two versions of the twin tasks do not match, they will save the task results of the first version and the second version, wait for the third version of the task to finish execution, judge the correct execution result of the task, and finally restore the state of the faulty task stack to the initial state to prepare for the next task to run. The I / O tasks can be in two situations: normal operation and failure. First, the manager tasks execute the input acceptance phase, then create two redundant versions of the I / O tasks accordingly and give the input, and at the same time set the execution time of these two twin tasks, where the execution time is equal to the sum of the longest execution times of the first and second twin tasks.

[0167] Since the priority of the first version of the twin tasks is higher than that of the second version of the twin tasks, the first version of the twin tasks must be executed first, and then the second version of the twin tasks is run. After both tasks are executed or the set execution time arrives, the manager tasks preempt the CPU resources and compare the execution results of these two tasks. If the results are equal, the manager tasks enter the I / O transfer phase.

[0168] If a transient failure occurs due to a soft error when the first version or the second version of the twin tasks is running, the manager tasks will detect the transient failure during the task state comparison and create the third version of the twin tasks. The priority of the third version of the twin tasks is higher than that of the manager tasks, so when the third version of the twin tasks finishes execution, the manager tasks will preempt the CPU resources and enter the voting phase. In the voting phase, if the results of two of the tasks are the same, the results of this type of task will be used for I / O transfer, otherwise the three twin tasks will be suspended to complete specific operations for subsequent exception handling.

[0169] 7. Reinforcement Learning-based I / O Interface Repair Method

[0170] Based on the overall reliability assessment of the I / O interface hardware system (IOHS) of the intelligent mobile system, a reinforcement learning-based intelligent mobile system I / O interface repair method is used to repair the I / O interface of the intelligent mobile system in a failed state.

[0171] An I / O interface fault tolerance device for an intelligent mobile system provided by the present invention has the functions of the above reliability analysis method and can perform fault tolerance repair according to the reliability analysis results. The functions of the device can be implemented by hardware or by software with corresponding functions. The device may include: an extraction module, a model construction and conversion module, a reliability evaluation module, and a fault tolerance module.

[0172] The extraction module is used to monitor the states of each functional module of the I / O interface of the intelligent mobile system, extract the hardware structure information in the IOHS of the intelligent mobile system, and obtain the state information of the functional modules; the state information includes the data saved in the receive data register, the data saved in the transmit data register, the control bit information of the control register, the status bit information and importance information in the status register.

[0173] The model construction and conversion module is used to construct a reliability model RMIO for the I / O interface functional module of the intelligent mobile system IOFM , and convert the RMIO IOFM model into a CTMC IO , and perform model element mapping through a pre-designed element mapping rule to generate a continuous-time Markov model.

[0174] The reliability evaluation module is used to evaluate the overall reliability of the IOHS according to the CTMC generated by the conversion, and calculate the states of the overall state space of the IOHS by using the relationship between states and the reliability evaluation interface of the functional modules. IO

[0175] The fault tolerance module is used to repair the I / O interface of the intelligent mobile system in a failed state by using an I / O interface repair method based on reinforcement learning according to the overall reliability evaluation result of the IOHS.

[0176] Specifically, the model construction and conversion module is used for:

[0177] Establish a reliability model RMIO according to the information obtained by the extraction module IOFM , and since the evolution process of the IOHS state over time can be described as a continuous-time Markov chain, perform model conversion according to the described mapping rule to generate a time-related CTMC IO model.

[0178] The I / O interface repair method based on reinforcement learning used by the fault tolerance module is constructed by the following method:

[0179] First, a unique code needs to be created for all the state distributions of the IOHS. This code should not only be able to distinguish all the states of the IOHS but also be associated with the reliability of the IOHS. For each state of the IOHS, there is a corresponding unique code, and this unique code is composed of the states of all the IOFMs in the IOHS. For example, Figure 3 the code for the state "RDR·SDR·CR·SR" in

[0180] is "RDR·SDR·CR·SR", indicating that RDR, SDR, CR, and SR are all in the normal state. Therefore, the state sequence of all the IOFMs is used as the state code of the IOHS. t ∈S, the operation that the IOHS can select from the set of allowed repair operations at the current moment is denoted as M t ∈M(S t ). After taking the repair operation, the state of the IOHS at the next moment is S t+1 ∈S + , and the reliability repair value brought by this repair operation is where all the states are denoted as S + , and the non - termination states are denoted as S.

[0181] Third, the characteristic equation obtained according to the Markov decision process is shown as the following formula, where p(s′,r|s,a) is the probability of transitioning to state s′ when considering the reliability repair value r and taking the repair operation m∈M(S) in state s.

[0182] p(s′,r|s,m) = p r [S t+1 = s′,R t+1 = r|S 0 ,M 0 ,S 1 ,M 1 ,...,S t ,M t

[0183] The decision - making process DP is defined as the five - tuple shown in the following formula, where S represents the state space; M represents the set of repair operations, and this set is finite; P represents the state transition probability from the initial state to the target state; R represents the reliability repair value of the IOHS obtained after taking the repair operation; γ∈[0,1] represents the attenuation rate, which quantifies the difference in reliability repair values.

[0184] ​DP = (S, M, P, R, γ)

[0185] In the fourth step, the value function of the repair model is shown as follows, where γ ∈ [0, 1] represents the attenuation rate, and R(s, m) is the reward function, that is, for state s IO taking the repair operation m IO brings the reward. Multiply the discount factor and the reward function and sum over the repair steps, which is the final objective function. Select the repair step that can produce the maximum objective function. By making the best action at each state, the final repair step can be obtained

[0186]

[0187] In the fifth step, take all states of IOHS as the state transition set, and set the failure state of IOHS as the initial state. Both instantaneous faults and permanent faults may cause the IOFM in IOHS to become the failure state. For the IOFM failure caused by instantaneous faults, it can be repaired by fault-tolerant methods such as refresh reconfiguration and wait rewrite. For the IOFM failure caused by permanent faults, methods such as hardware-implemented multi-mode redundancy and component replacement can be used for repair. In order to make the overall I / O interface of the intelligent mobile system return to the normal working state in the shortest time, use the ε-greedy algorithm to calculate and take the result of the above formula and take the repair action that can obtain the maximum value. Select the IOFM to be repaired from the state transition set and take the optimal action to repair it, and then repeat this repair action until IOHS returns to the normal state

[0188] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed< / iofmname> < / iofmname> < / iofmname>

Claims

1. A method for analyzing the reliability of the I / O interface of an intelligent mobile system, characterized in that, the method comprises the following steps: Extract the reliability-related information in the design of the I / O interface of the intelligent mobile system, perform hierarchical modeling, and establish a reliability model RMIO for the I / O interface of the intelligent mobile system from bottom to top in units of the I / O interface function modules of the intelligent mobile system; by extracting the hardware structure information in the I / O interface hardware system IOHS of the intelligent mobile system, divide the IOHS into different I / O interface function modules of the intelligent mobile system, namely IOFM, and the IOFM includes a receive data register RDR, a send data register SDR, a control register CR, and a status register SR; Describe the RMIO model using a Continuous-Time Markov Chain (CTMC), establish the mapping relationship, and construct the CTMC IO model, and establish a CTMC IO model for each intelligent mobile system I / O interface function module (IOFM); The Markov Chain Monte Carlo algorithm MCMC is used to randomly sample the failure rate of the I / O interface functional modules of the intelligent mobile system, and based on the CTMC IO evaluate the reliability of each functional module of the I / O interface of the intelligent mobile system, and calculate the instantaneous availability of each functional module of the I / O interface of the intelligent mobile system based on this; Describe the evolution process of the I / O interface hardware system (IOHS) state of the intelligent mobile system over time as a continuous-time Markov chain (CTMC). IO , and complete the reliability assessment of the overall IOHS based on this Markov chain; the normal or failed states of all functional modules constitute the state space S of the overall IOHS IO , S IO The state transition probability matrix A in IO = [a ij is calculated through the relationships between states and the reliability assessment results of functional modules; According to the overall reliability evaluation result of the IOHS, use an I / O interface fault tolerance scheme based on task redundancy to reinforce the IOFM with reliability lower than the set threshold, generate a manager task to create twin I / O tasks, and compare the execution results of the twin I / O tasks to complete fault detection; On the basis of the overall reliability evaluation of the I / O interface hardware system IOHS of the intelligent mobile system, use an I / O interface repair method based on reinforcement learning to repair the I / O interface of the intelligent mobile system in a failure state to restore it to a normal state in the shortest time; first, take the failure state as the initial state in the state space, formulate a deterministic policy and a reward function, and finally use the ε-greedy algorithm to determine the optimal policy, and select the IOFM for repair according to the optimal policy.

2. The method for analyzing the reliability of the I / O interface of an intelligent mobile system according to claim 1, characterized in that, The establishment of the reliability model of the I / O interface function module of the intelligent mobile system is expressed as the following triple: RMIO IOFM =(FR IO ,SS IO ,STR IO ) In the formula, RMIO IOFM is the functional reliability model of the I / O interface of the intelligent mobile system, and FR IO represents the failure rate of the I / O interface function module IOFM of the intelligent mobile system; SS IO represents the state space of IOFM; STR IO represents the state transition relationship, where STR IO includes the source state sStateIO and the target state tStateIO; the failure rates of the receive data register RDR, the transmit data register SDR, the control register CR, and the status register SR of IOFM are λ RDR , λ SDR , λ CR , λ SR , respectively. Let the importance weights of IOFM in different intelligent mobile systems be ω RDR , ω SDR , ω CR , ω SR , respectively. The calculation formula for the failure rate λ IOFM of IOFM is: λ IOFM = 1 - Π i∈(RDR,SDR,CR,SR) (1 - ω i λ i ).

3. The method for analyzing the reliability of the I / O interface of an intelligent mobile system according to claim 1, characterized in that, Continuous-Time Markov Chain (CTMC) IO is a five-tuple: CTMC 10 = (S IO , S in , A IO , T IO , t) Among them, S IO represents the state space of the CTMC IO , S in ∈S IO represents the initial state of the CTMC IO , A IO = [a ij represents the state transition probability matrix, a ij represents the probability of transitioning from state S i ∈S IO to state S j ∈S IO ; is the set of state transition relations, (S i , S j ) represents the existence of a transition from state S i to state S j ; t represents time.

4. The method for analyzing the reliability of the I / O interface of an intelligent mobile system according to claim 3, characterized in that, Describe RMIO as a CTMC IO and perform the conversion by designing the element mapping rules between RMIO and CTMC IO The mapping rules are shown in Table 1 below: Table 1 RMIO Modeling and CTMC IO Mapping Rules between Elements 5. The method for analyzing the reliability of the I / O interface of an intelligent mobile system according to claim 2, characterized in that, The failure rate λ of the IOFM IOFM has the following functional relationship with time: The typical state transition matrix of the function module of the I / O interface of the intelligent mobile system is: The parameter meanings are shown in Table 2 below: Table 2 Parameter of the IOFM State Transition Relationship Diagram From this, the state probability equation of the I / O interface function module of the intelligent mobile system is obtained as: Among them, P N (t), P R (t), P DG (t), P D (t), P F (t), P E (t) respectively represent the probabilities that the IOFM is in the normal state, recovery state, degraded error-tolerant working state, fault detection state, failure state, and fault state at time t, and t′ is the next moment after time t; Reliability R of the I / O interface function module IOFM (t) The evaluation calculation formula is as follows: R IOFM R(t) = 1 - P E (t) Among them, R IOFM (t) refers to the probability of not being in the failure state FailureState at time t.

6. The method for analyzing the reliability of the I / O interface of an intelligent mobile system according to claim 1, characterized in that, When evaluating the reliability of the I / O interface hardware system of an intelligent mobile system, the state space of the overall I / O interface of the intelligent mobile system is composed of the states of all functional modules, that is, a state sequence s of all functional modules n-1 s n-2... s 1 s 0 constitutes a state of the overall I / O interface of the intelligent mobile system, where s j represents that the functional module IOFM j is in the normal state s j = 0 or the failure state s j = 1, 0 ≤ j < n, where n is the number of functional modules.

7. The method for analyzing the reliability of the I / O interface of an intelligent mobile system according to claim 1, characterized in that, When the system is in an abnormal working state, use a repair method based on reinforcement learning to repair the I / O interface hardware system IOHS of the intelligent mobile system in a failure state.

8. An I / O interface fault tolerance device for an intelligent mobile system based on the method according to any one of claims 1 to 7, characterized in that, the device comprises: An extraction module, configured to monitor the status of each functional module of the I / O interface of the intelligent mobile system, extract the hardware structure information in the IOHS of the intelligent mobile system, and obtain the status information of the functional modules; the status information includes the data stored in the receive data register, the data stored in the transmit data register, the control bit information of the control register, the status bit information in the status register, and the importance information; Model construction and conversion module, used to construct the reliability model RMIO of the I / O interface function module of the intelligent mobile system IOFM , convert the RMIO IOFM model into a CTMC IO , and perform model element mapping through pre-designed element mapping rules to generate a continuous-time Markov model; A reliability evaluation module for evaluating the overall reliability of the IOHS based on the CTMC generated by the transformation IO calculate the states of the overall state space of the IOHS using the relationships between the states and the reliability evaluation interfaces of the functional modules A fault tolerance module, configured to repair the I / O interface of the intelligent mobile system in a failed state according to the overall reliability evaluation result of the IOHS, using an I / O interface error detection scheme based on task redundancy and an I / O interface repair method based on reinforcement learning.

9. The I / O interface fault tolerance device of the intelligent mobile system according to claim 8, characterized in that the fault tolerance module creates a manager task using an I / O interface error detection method based on task redundancy and generates two twin versions of the I / O task by the manager task; the manager task is responsible for the input preparation, scheduling, output comparison, and error recovery operations of the twin tasks; first, the first version and the second version of the twin tasks are run. If the execution results of these two twin tasks are different, the third version of the twin task will be run. Once the manager task detects that the outputs of the first two versions of the twin tasks do not match, it will save the task results of the first version and the second version, wait for the third version of the task to execute, determine the correct execution result of the task, and finally restore the status of the fault task stack to the initial state to prepare for the operation of the next task.

10. The I / O interface fault tolerance device of the intelligent mobile system according to claim 9, characterized in that the I / O interface repair method based on reinforcement learning used by the fault tolerance module is constructed by the following method: The first step is to create a unique encoding for all state distributions of the IOHS. This encoding can not only be used to distinguish all states of the IOHS, but also be associated with the reliability of the IOHS; for each state of the IOHS, there is a unique encoding corresponding to it, and this unique encoding is composed of the states of all IOFMs in the IOHS. The state sequence of all IOFMs is used as the state encoding of the IOHS. Second step, at time t, the IOHS receives the state of the external environment as S t ∈S, the operation that the IOHS can select from the set of allowable repair operations at the current state at the current moment is denoted as M t ∈M(S t ), after taking the repair operation, the state of the IOHS at the next moment is S t+1 ∈S + , and the reliability repair value brought by this repair operation is Among them, all states are denoted as S + , and the non-terminal states are denoted as S; The third step, the characteristic equation obtained according to the Markov decision process is as follows: p(s′, r|s, m) = p r [S t+1 = s′, R t+1 = r|S 0 , M 0 , S 1 , M 1 ,..., S t , M t ​ Among them, \(p(s′,r|s,m)\) is the probability of transitioning to state \(s′\) when considering the reliability repair value \(r\) in state \(s\) and taking the repair operation \(m\in M(S)\); this probability is determined by the state of the IOHS and the set of repair operations \(S\) 0 , \(M\) 0 , \(S\) 1 , \(M\) 1 ,..., \(S\) t , \(M\) t \(\in(S,M(S))\) represents the conditional probability that the state at time \(t + 1\) is \(s\) and the reliability repair value at time \(t + 1\) is \(r\), where the subscript \(r\) represents the reliability repair value; The decision-making process DP is defined as the five-tuple shown in the following formula, DP=(S,M,P,R,γ) where S represents the state space; M represents the set of repair operations, and this set is finite; P represents the state transition probability from the initial state to the target state; R represents the IOHS reliability repair value obtained after taking the repair operation; γ∈[0,1] represents the attenuation rate, which quantifies the difference in the reliability repair value; The fourth step, the value function of the repair model is as follows: where R(s IO (k), m IO (k)) is the reward function, that is, for the state s IO (k), taking the repair action m IO (k) brings the reward, and γ k is the decay rate of the cumulative reliability repair value from the initial time to time k; multiplying the discount factor and the reward function and summing over the repair steps gives the final objective function. Select the repair step that can produce the maximum objective function. By making the best action at each state, the final repair step is obtained; Step 5: Use all the states of IOHS as the state transition set, and set the failure state of IOHS as the initial state; for the failure of IOFM caused by transient faults, repair it through fault-tolerant methods such as refreshing reconfiguration and waiting for rewriting. For the failure of IOFM caused by permanent faults, use methods such as hardware-implemented multi-mode redundancy and component replacement for repair; use the ε-greedy algorithm to calculate and adopt the result of the formula in Step 4 to obtain the repair action with the maximum value, select the IOFM to be repaired from the state transition set and take the optimal action to repair it, and then repeat this repair action until the IOHS returns to the normal state.

Images