Transformation method for digitizing vehicle function design to realize self-adaption

Abstract

The invention relates to the field of vehicle function design, and is applied to a design mode for transforming a traditional function, so that the development amount of the function is reduced, and meanwhile, the method can be suitable for AI chip operation to realize a more intelligent behavior. Intelligent essence is personification, state recognition and demand expression of human beings not only use one threshold value, but also do not adopt A to B single selection, and more are compromised and variable, vehicle body function design outputs authenticity of each condition through a single threshold value, then a plurality of condition decisions are combined for activation, one of loads is selected for execution when a plurality of functions are jointly activated, and the vehicle body function design is more accurate. The characteristics are against habits of people and are difficult to give tacit understanding to users, in addition, the traditional functions of full logic design cannot utilize digital bonus to get rid of a development mode that the functions are changed only by trial and error, and the developed functions are static and cannot adapt to diversity. According to the invention, the rule is digitalized, and the switching value is linearized, so that the operation and self-adaption are realized, and an operation scheme is provided.

Description

Technical Field

[0001] The present invention relates to the field of vehicle function design, and is applied to the design method of transforming traditional functions so that not only the development workload of the functions is reduced, but also the functions become suitable for AI chip operation to realize more intelligent behaviors. Background Art

[0002] The essence of intelligence is anthropomorphism. Humans do not use only one threshold to identify states and express needs, and the expected execution is not a single choice of A or B, but more of a compromise and variable. Vehicles, especially body function design, have always used only a single threshold to output the authenticity of conditions, and several conditions are taken to activate decision-making functions. When multiple functions are activated together, the load selects one to execute. These characteristics are contrary to people's habits, which is why it is difficult to give users tacit understanding even if the functions are expanded. In addition, traditional functions based on full logic design cannot use the new digital productivity to get rid of the development mode that relies solely on trial and error. The functions developed are static and cannot adapt to diverse needs. Summary of the invention

[0003] To this end, the present invention proposes the concept of digitizing vehicle function design and provides a systematic solution to transform the root fully logical design into a linear mathematical operation model as a technical foundation for further realizing self-adaptation.

[0004] In order to achieve the above object, the present invention provides the following technical solutions.

[0005] A method for linearizing vehicle function design to achieve self-adaptation includes a linearization architecture scheme, an expression method for judging conditions, a calculation method for function activation, and a linearization execution strategy. First, the conditions that originally corresponded only to signals one by one are changed into subjective variables close to human behavior, and the function activation intensity corresponding to the conditions is obtained through specific mathematical operations. Then, the execution intensity of the load required by the function is obtained through specific operations to complete a driving instruction. In order to achieve self-adaptation, a coefficient matrix that can be automatically adjusted by further strategy implementation is introduced into the calculation process.

[0006] The expression method of the judgment condition adopts a subjective evaluation value to replace the direct one-to-one expression of the condition and the signal, and associates the signal with the subjective condition in a many-to-one manner, and then uses a set set to fuzzy the signal value to obtain the membership, and obtains a numerical output between 0 and 1 to express the corresponding maturity of the condition, changing the original form of 0 or 1 that only indicates whether it is established or not.

[0007] The calculation method for function activation is to further perform weighted summation of the obtained conditional values ​​according to the importance of the relevant functions and then perform activation processing to output a numerical value between 0 and 1 to express the demand intensity of the function.

[0008] The described linearization execution strategy is to further set the request intensity for each required function pointing to the actuator through analog fuzzy inference calculation to obtain a numerical value transmission from -1 to 1, reflecting the superimposed or compromised driving force.

[0009] Therefore, the described method at least includes the calculation of specific condition membership degrees, the method of function activation operation, the calculation of linear execution driving force, and a system framework solution including the adjustment of the adaptive coefficient matrix.

[0010] Compared with the existing function operation mode, the present invention has the following advantages.

[0011] The transformation method of the present invention uses digital means to solve the performance required by the whole vehicle at each moment. Compared with the original logic-based operation mode, it can make the vehicle behavior evolve from rigid zero-sum execution to dynamic compromise operation, and the behavior performance is more anthropomorphic.

[0012] The transformation method of the present invention uses matrix operation mode, which can update the whole function performance at one step by means of increment and parameter import, solve the inefficient development of implementing functions one by one in code and repeatedly trial-and-error modification due to conflicts, and save a large amount of costs.

[0013] The transformation method of the present invention uses the digital model method. Not only does it use model operation to replace the judgment of static rules and improve the design and operation quality of functions, but it is also an essential measure to liberate functions from the constraint of only being able to run on logic-based MCUs and make them run on AI digital processing chips.

[0014] The inspection method of the present invention uses digital matrix operation to solve execution instructions, which is also beneficial to intelligent exploration, and the parameterized model structure better supports adaptive closed-loop adjustment.

[0015] The additional description of the present invention will be given in part in the following description, and its advantages will become obvious from the following description.

[0016] Brief description of the drawings. The above advantages of the present invention will become obvious and easy to understand from the following combination of drawings for inspection requirements and implementation examples, where.

[0017] Figure 1 It is the overall architecture solution for function digitization.

[0018] Figure 2 It is the schematic diagram of the calculation process for function digitization.

[0019] Detailed implementation manners. Embodiments of the present invention are described in detail herein. Examples of the embodiments are shown in the accompanying drawings. The embodiments described by referring to the accompanying drawings are exemplary and are only used to explain the present invention and shall not be construed as a limitation to the present invention.

[0020] As Figure 1 shown, the architecture solution of the present invention includes conditional linear output calculation 1 based on signal states, functional linearized output calculation 2 based on conditions, execution linearized output calculation 3 based on functions, and operation with parameter adaptation 4. In addition, an influence coefficient matrix 10 reflecting the influence of each state on condition judgment, a weight coefficient matrix 20 reflecting the weight of each condition on function activation, and a strength coefficient matrix 30 reflecting the strength of each function on execution request are set as components for full digital operation from state reading to actuator drive. According to this framework structure, the situation of continuously piling up logics can be jumped out. After setting each coefficient table, the execution amount can be obtained only by means of mathematical operations. By continuously reading -> calculating -> outputting the matrix, the output value of the actuator at each moment can be refreshed.

[0021] As Figure 1 shown, for the conditional linear output calculation based on signal states, the implementation process includes selecting a unified fuzzy set to describe the condition, selecting the signal states affecting this condition, presetting the influence coefficient w of each state, dividing the state reading value by the discriminant quantity to obtain the input to be solved, fuzzyfying the input to obtain the membership degree and filtering by respective thresholds, and multiplying the processed membership degrees by the coefficients and summing to obtain the condition maturity output.

[0022] Figure 1 shown, for the functional digital output calculation based on conditions, the process includes selecting the conditions that can trigger function execution, presetting the weighting coefficient r and activation threshold of each condition, and filtering the condition input value through weighted summation and threshold to obtain the activation degree, that is, the function demand intensity.

[0023] Figure 1 shown, for the execution linearized output calculation based on functions, the process includes presetting the linear relationship between the demand intensity of each function and the strength of the actuator, amplifying the demand intensity by respective strength coefficients n and then calculating the combined request strength, and obtaining the strength output of the execution action through threshold processing.

[0024] Further, as Figure 2 shown, three intermediate vectors are constructed. The condition maturity vector 100 is obtained through the state input vector via matrix 10 and specified calculation. Then, the function demand intensity vector 200 is obtained from the condition maturity vector 100 via matrix 20 and specified calculation. And the execution vector strength vector 300 is obtained by multiplying the function demand intensity vector 200 by matrix 30. These are the outputs of each calculation link.

[0025] Further, asFigure 2 For the calculation steps of the condition maturity vector 100 shown, denote the state vector as x and the corresponding discrimination vector as y. Calculate (x + 1) / (y + 1) -> obtain values -> vector screening -> threshold filtering -> multiplication and summation of influence coefficients -> normalization processing and output. The advantage of using division is that it is convenient to uniformly process the on-off values, analog values, and enumerated values that are common in state signals. The descriptions of the various operation components used are listed in the figure.

[0026] Furthermore, for the calculation steps of the functional requirement intensity vector 200, multiply the vector 100 by the weighting matrix 20 -> threshold filtering -> directly output.

[0027] Furthermore, for the calculation steps of the execution vector strength vector 300, denote the output of function 1 as a1 and the output of function 2 as a2. Multiply the vector 200 by the matrix 30 to obtain the combined strength -> calculate the AB direction -> select the threshold -> threshold filtering -> output the combined strength -> look up the corresponding output value based on the two-way demand-strength mapping table. The specific descriptions of the operation components are listed in the figure.

[0028] Furthermore, for Figure 1 the implementation strategy of the parameter adaptive operation shown, for each execution operation, collect and compare the deviation between the running value of this actuator and the strength output value within the specified time, and use it as the input for reverse adjustment of the parameter matrices w, r, and n. The specific method is not described here.

Claims

1. A method for digitalizing vehicle function design to achieve self-adaptation, characterized in that: A digital architecture scheme, an expression of judgment conditions, a calculation method for function activation, and a linear execution strategy. The digital architecture scheme is used to propose that the switch-type function judgment execution process can be replaced by a digital concept and a system implementation framework. The expression of judgment conditions is used to propose how to transform the truth value expression into a linear one. The calculation method for function activation is used to propose how to change the logical truth value judgment of function activation into a digital operation intensity calculation and calculate the function activation intensity value. The linear execution strategy is used to propose how to select the execution strategy for the digital function trigger conditions.

2. The method for realizing self-adaptation by digitalizing vehicle function design according to claim 1, characterized in that: The digital architecture solution proposes to transform the switch-type function activation and load execution methods into a concept that can linearly judge the maturity of conditions, linearly express the intensity of function activation requirements, and linear execution actions that are compatible with multi-function requests, as well as a specific implementation system framework.

3. The method for realizing self-adaptation by digitalizing vehicle function design according to claim 1, characterized in that: The expression method of the judgment condition proposes a method of converting a single threshold value into a linear output reflecting maturity by comparing the logical value, and the proposed implementation method.

4. The method for realizing self-adaptation by digitalizing vehicle function design according to claim 1, characterized in that: The input conditions and activation states proposed in the calculation method for function activation are transformed from logical values ​​into calculation methods using digital quantities, as well as the proposed implementation principles.

5. The method for realizing self-adaptation by digitalizing vehicle function design according to claim 3, characterized in that: This calculation method uses fuzzy processing of associated states and weighted calculation of the linear maturity of the corresponding conditions.

6. The method for realizing self-adaptation by digitalizing vehicle function design according to claim 1, characterized in that: This linearized execution strategy adopts the demand intensity value, transforms the switch-style exclusive operation into a compromise operation for multiple requests, and the proposed implementation strategy.

7. The method for implementing digital execution of vehicle functions according to claim 2, characterized in that: In the design framework, all serial calculation output links are equipped with coefficient tables that can perform matrix operations and the proposed usage methods.

8. The method for digitalizing vehicle function design to achieve self-adaptation according to claim 7, characterized in that: A feedback loop is used in the design framework, which uses the collected user intervention data on the executed actions to calculate the expected deviation to close the loop and correct the coefficient table in operation or training to achieve the adaptive goal.

Images