An adaptive control method for AMT shifting based on driving intention and environment

Abstract

The present invention discloses an adaptive control method for automatic transmission (AMT) shifting based on driving intention and environment, comprising the following steps: (1) establishing a Bolza optimal control model for the shifting stage, and realizing the shifting requirements under different driving intentions by allocating weight factors; (2) solving the Bolza optimization model by pseudo-spectral method and quadratic programming, and obtaining the optimal speed change rates of the engine and clutch under three different driving intentions; (3) performing offline training by using a BP neural network and a rule table, integrating the change rate signal output by the optimal speed call layer by an integrator of a preprocessing layer, and obtaining a reference speed curve for subsequent controller calling; (4) adopting a speed closed-loop scheme, designing an adaptive sliding mode controller for the engine / clutch during the shifting process, and using the engine / clutch reference trajectory obtained after integration as a target trajectory, which is input into the shifting model for tracking control; (5) verifying different driving intentions and environments in Matlab / Simulink; the present invention realizes an AMT shifting process with the best comprehensive multi-objective evaluation index, and improves the control speed of the AMT vehicle shifting process.

Description

Technical Field

[0001] The present invention relates to the technical field of clutch adaptive control, and in particular to an AMT shift adaptive control method based on driving intention and environment. Background Art

[0002] In actual engineering applications, the clutch shift engagement trajectory is usually determined by table lookup. However, the table lookup method is difficult to adapt to changes in the driver's subjective driving intention during driving. AMT engineering vehicles usually use PID control or open-loop control strategies in actual shift control. These strategies rely on offline calibrated parameters to implement the control process. However, due to the complexity of the actual operating environment, such as vehicle body mass, road conditions and time-varying disturbances, traditional control strategies cannot meet the needs of precise control. Therefore, in order to improve the accuracy of the AMT vehicle shift process, it is necessary to formulate an AMT vehicle shift process target engagement trajectory that conforms to the driver's intention and an adaptive control strategy that meets the changing external environment.

[0003] Current research on clutch shift control can be categorized into two main types: optimization-based control and target-engaged trajectory tracking. Optimization-based shift control employs a variety of optimization algorithms, including model predictive control, linear quadratic optimal control, and sequential quadratic programming. Control methods based on shift reference trajectory tracking determine the optimal reference trajectory for the shift process at the upper level of the system and design a shift controller to track this trajectory, thereby improving shift quality. In recent years, numerous researchers have conducted extensive research on shift control for automated manual transmission (AMTs). A key focus is on achieving rapid clutch control response and high-precision performance to adapt to varying driver intentions and driving environments, ensuring smooth and efficient vehicle shifts. Research on shift control for engineering vehicles typically relies on table lookup methods to determine the target clutch engagement trajectory, with limited consideration of the comprehensive optimization of shift process evaluation metrics. This approach suffers from insufficient adaptability and control accuracy in the face of significant external environmental disturbances, resulting in poor robustness of the control system. Therefore, improving the model's adaptability to environmental disturbances and control accuracy while maintaining computational efficiency remains a key challenge in shift control research. Summary of the Invention

[0004] Purpose of the invention: The purpose of the present invention is to provide an AMT shift adaptive control method based on driving intention and environment to solve the problems existing in the background technology.

[0005] Technical solution: The present invention provides an AMT shift adaptive control method based on driving intention and environment, comprising the following steps:

[0006] (1) Establish the Bolza optimal control model during the gear shifting phase and realize the gear shifting requirements under different driving intentions by allocating weight factors;

[0007] (2) The Bolza optimization model is solved by pseudospectral method and quadratic programming to obtain the optimal speed change rates of the engine and clutch under three different driving intentions;

[0008] (3) Offline training is performed using the BP neural network and the rule table. The rate of change signal output by the optimal speed call layer is integrated through the integrator of the preprocessing layer to obtain a reference speed curve for subsequent controller calls.

[0009] (4) A speed closed-loop scheme is adopted to design an adaptive sliding mode controller for the engine / clutch during the shifting process. The engine / clutch reference trajectory obtained after integration is used as the target trajectory and input into the shifting model for tracking control.

[0010] (5) Verify different driving intentions and environments in Matlab / Simulink.

[0011] Furthermore, in step (1), the dynamic model of the shifting phase is as follows:

[0012]

[0013] Among them, T e T is the torque acting on the clutch active plate from the engine output end; c is the clutch torque; T f is the driving resistance torque equivalent to the driven disc; ω e 、ω c I is the clutch master and driven disc speed; e , I c is the equivalent moment of inertia of the driving and driven ends; b1 and b2 are the rotational damping coefficients of the engine and clutch;

[0014] The Bolza optimal control model formula during the shifting phase is as follows:

[0015]

[0016] Among them, the first constraint is the state differential equation, the second constraint is the boundary condition, and the third constraint is the path restriction. The state and control variables to be solved are discretized at the LGR point and converted into approximate continuous variables through Lagrange interpolation. The optimal control problem is converted into a nonlinear programming problem, and the optimal solution is obtained through iteration.

[0017] Furthermore, step (2) includes the following steps:

[0018] (21) System time domain transformation: Assume that the time domain space of the original optimization problem is discretized into n+1 nodes, and the time domain interval between the k-1th node and the kth node is defined as [t k-1 ,t k ], k=1,2,…,n, and t0<t1<…<t n =t f The Lengendre orthogonal polynomials are defined as intervals [-1, 1], then the k-th time domain interval is transformed into the original time domain t∈[t k-1 ,t k ] is converted to τ∈[-1,1], the formula is as follows:

[0019]

[0020] Derivative the variable τ in the new time domain with respect to the variable t in the original time domain yields:

[0021]

[0022] According to equations (5.8) and (5.9), the original optimization problem is redefined as:

[0023]

[0024] (22) LGR collocation and Lagrange interpolation polynomial approximation: discretize the k-th subinterval after time domain transformation into N LGR collocation points With a non-matching point The state quantity is expressed as Construct N+1 order Lagrange interpolation polynomial

[0025]

[0026] Using the Lagrange interpolation polynomial as the basis function and combining the state variables corresponding to the non-collocated points in the interval after the time domain transformation, the state variables and control variables in the original optimization problem can be approximated as follows:

[0027]

[0028] (23) Convert the Bolza problem into a nonlinear programming problem: Use multi-segment interval LGR integration to approximate the integral term in the objective function, and the objective function can be converted into:

[0029]

[0030] in, is the LGR integral weight, which is expressed as

[0031] The state variables in the approximated formula (5.12) are derived to obtain:

[0032]

[0033] Where: is the derivative of the interpolation polynomial, which is expressed as

[0034] The equality constraints at the endpoints of the interval and the inequality constraints within the interval can be approximated as:

[0035]

[0036] In summary, the Bolza problem of the AMT shifting process can be transformed into a nonlinear programming problem:

[0037]

[0038] (24) The converted nonlinear programming problem is solved using a sequential quadratic programming algorithm; specifically, the initial point is selected as the starting point of the problem, and the iteration limit or convergence criterion is set. Then, the Lagrange multiplier function is introduced to transform the original nonlinear programming problem into a constrained optimization problem. At the current point, the objective function and constraints are linearized. Taylor expansion or approximate linearization method is usually used to obtain an approximate Hessian matrix. Based on the linearized objective function and constraints, a quadratic programming subproblem is solved to obtain the corresponding search direction. The iteration is continued until convergence to obtain the optimal solution.

[0039] Furthermore, in step (3), the engine reference speed calling rule is as follows: the three parameters of accelerator pedal opening, initial acceleration and shift time are used as training feature parameters of the neural network, and the rate of change of the target speed obtained by pseudo-spectral method and quadratic programming is used as label data. When calling, the output of the neural network model is directly integrated to obtain the engine reference speed.

[0040] Furthermore, in step (3), the clutch reference speed calling rule is as follows: different clutch speed change rate values ​​are given at different throttle openings, and the clutch speed measured by the sensor at the start of the shift is used as the initial value of the reference speed. The reference speed change rate at different throttle openings and the clutch speed at the start of the shift are saved offline in the rule table. The initial value is integrated during the call to obtain the reference speed of the clutch during the shift process.

[0041] Furthermore, in step (4), the state equation of the clutch adaptive sliding mode controller during the shifting process is designed as follows:

[0042]

[0043]

[0044] Where, d f (t) is the unknown resistance disturbance during vehicle driving, which can be regarded as the change of vehicle driving resistance caused by external environment such as road surface and slope during driving.

[0045] Furthermore, in step (4), the sliding mode surface of the clutch adaptive sliding mode controller during the shifting process is designed as follows:

[0046]

[0047]

[0048] Among them, c c >0 is the sliding surface coefficient.

[0049] Furthermore, in step (4), the Lyapunov stability of the clutch adaptive sliding mode controller during the shifting process is proved as follows:

[0050]

[0051] Among them, ε cl and p c is the control parameter,

[0052] Furthermore, in step (4), the state equation of the engine adaptive sliding mode controller during the shifting process is designed as follows:

[0053]

[0054] Furthermore, in step (4), the sliding mode surface of the engine adaptive sliding mode controller during the shifting process is designed as follows:

[0055]

[0056] The Lyapunov stability of the engine adaptive sliding mode controller during the shifting process is proved as follows:

[0057]

[0058] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: a shifting process optimization model that comprehensively considers the shifting shock degree, sliding friction work and shifting time is established, and the shifting requirements under different driving intentions are realized by allocating weight factors. The optimal trajectory of the shifting process is obtained by pseudo-spectral method and quadratic programming, and is called using a rule table and a trained BP neural network respectively. An adaptive sliding mode controller for the shifting process of an AMT vehicle is designed to track the optimal reference speed of the clutch called online. An engine controller is designed to track the optimal reference speed of the engine by adjusting the throttle opening; an AMT shifting process with comprehensive optimization of multi-objective evaluation indicators is realized; and the control speed of the shifting process of an AMT vehicle is improved by calling and tracking the optimal reference trajectory. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0060] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0061] like Figure 1 As shown, an embodiment of the present invention provides an AMT shift adaptive control method based on driving intention and environment, comprising the following steps:

[0062] (1) The Bolza optimal control model for the shifting phase is established, and the shifting requirements under different driving intentions are realized by allocating weight factors. The dynamic model of the shifting phase is as follows:

[0063]

[0064] Among them, T e T is the torque acting on the clutch active plate from the engine output end; c is the clutch torque; T f is the driving resistance torque equivalent to the driven disc; ω e 、ω c I is the clutch master and driven disc speed; e , I c is the equivalent moment of inertia of the driving and driven ends; b1 and b2 are the rotational damping coefficients of the engine and clutch;

[0065] The Bolza optimal control model formula during the shifting phase is as follows:

[0066]

[0067] Among them, the first constraint condition is the state differential equation, the second constraint condition is the boundary condition, and the third constraint condition is the path restriction. The state and control variables to be solved are discretized at the LGR point and converted into approximate continuous variables through Lagrange interpolation. The optimal control problem is converted into a nonlinear programming problem, and the optimal solution is obtained through iteration.

[0068] The weight factor distribution under different driving intentions is shown in Table 1.

[0069] Table 1: Weight factor distribution table for different driving intentions

[0070]

[0071] (2) Solving the Bolza optimization model using pseudospectral method and quadratic programming to obtain the optimal speed change rates of the engine and clutch under three different driving intentions; the following steps are included:

[0072] (21) System time domain transformation: Assume that the time domain space of the original optimization problem is discretized into n+1 nodes, and the time domain interval between the k-1th node and the kth node is defined as [t k-1 ,t k ], k=1,2,…,n, and t0<t1<…<t n =t f The Lengendre orthogonal polynomials are defined as intervals [-1, 1], then the k-th time domain interval is transformed into the original time domain t∈[t k-1 ,t k ] is converted to τ∈[-1,1], the formula is as follows:

[0073]

[0074] Derivative the variable τ in the new time domain with respect to the variable t in the original time domain yields:

[0075]

[0076] According to equations (5.8) and (5.9), the original optimization problem is redefined as:

[0077]

[0078] (22) LGR collocation and Lagrange interpolation polynomial approximation: discretize the k-th subinterval after time domain transformation into N LGR collocation points With a non-matching point The state quantity is expressed as Construct N+1 order Lagrange interpolation polynomial

[0079]

[0080] Using the Lagrange interpolation polynomial as the basis function and combining the state variables corresponding to the non-collocated points in the interval after the time domain transformation, the state variables and control variables in the original optimization problem can be approximated as follows:

[0081]

[0082] (23) Convert the Bolza problem into a nonlinear programming problem: Use multi-segment interval LGR integration to approximate the integral term in the objective function, and the objective function can be converted into:

[0083]

[0084] in, is the LGR integral weight, which is expressed as

[0085] The state variables in the approximated formula (5.12) are derived to obtain:

[0086]

[0087] Where: is the derivative of the interpolation polynomial, which is expressed as

[0088] The equality constraints at the endpoints of the interval and the inequality constraints within the interval can be approximated as:

[0089]

[0090] In summary, the Bolza problem of the AMT shifting process can be transformed into a nonlinear programming problem:

[0091]

[0092] (24) Use the sequential quadratic programming algorithm to solve the transformed nonlinear programming problem; specifically, the following steps are performed: first, select the initial point as the starting point of the problem, and set the iteration limit or convergence criterion. Then, introduce the Lagrange multiplier function to transform the original nonlinear programming problem into a constrained optimization problem. At the current point, linearize the objective function and constraints. Usually, use Taylor expansion or approximate linearization method to obtain the approximate Hessian matrix. Based on the linearized objective function and constraints, solve a quadratic programming subproblem and obtain the corresponding search direction. Continue iterating until convergence to obtain the optimal solution.

[0093] (3) Offline training is performed using the BP neural network and the rule table. The rate of change signal output by the optimal speed call layer is integrated through the integrator of the preprocessing layer to obtain a reference speed curve for subsequent controller calls. The engine reference speed call rule is as follows: the three parameters of accelerator pedal opening, initial acceleration and shift time are used as training feature parameters of the neural network, and the rate of change of the target speed obtained by pseudo-spectral method and quadratic programming is used as label data. When calling, the output of the neural network model is directly integrated to obtain the engine reference speed.

[0094] The clutch reference speed calling rule is as follows: different clutch speed change rate values ​​are given at different throttle openings, and the clutch speed measured by the sensor at the start of the shift is used as the initial value of the reference speed. The reference speed change rate at different throttle openings and the clutch speed at the start of the shift are saved offline in the rule table. The initial value is integrated when calling to obtain the reference speed of the clutch during the shift process.

[0095] (4) Using a speed closed-loop scheme, an adaptive sliding mode controller for the engine / clutch in the shifting process is designed. The engine / clutch reference trajectory obtained after integration is used as the target trajectory and input into the shifting model for tracking control.

[0096] The state equation of the clutch adaptive sliding mode controller during the shifting process is designed as follows:

[0097]

[0098]

[0099] Where, d f (t) is the unknown resistance disturbance during vehicle driving, which can be regarded as the change of vehicle driving resistance caused by external environment such as road surface and slope during driving.

[0100] The sliding mode surface of the clutch adaptive sliding mode controller during the shifting process is designed as follows:

[0101]

[0102]

[0103] Among them, c c >0 is the sliding surface coefficient.

[0104] The Lyapunov stability of the clutch adaptive sliding mode controller during the shifting process is proved as follows:

[0105]

[0106] Among them, ε cl and p cFor the control parameters,

[0107] The state equation of the engine adaptive sliding mode controller during the shifting process is designed as follows:

[0108]

[0109]

[0110] The sliding mode surface of the engine adaptive sliding mode controller during the shifting process is designed as follows:

[0111]

[0112] The Lyapunov stability of the engine adaptive sliding mode controller during the shifting process is proved as follows:

[0113] (5) Verify different driving intentions and environments in Matlab / Simulink.

Claims

1. An adaptive control method for AMT shifting based on driving intention and environment, characterized in that: The following steps are involved: (1) The Bolza optimal control model for the shifting phase is established, and the shifting requirements under different driving intentions are realized by allocating weight factors. The dynamic model of the shifting phase is as follows: ； in, It is the torque acting on the clutch active plate from the engine output end; is the clutch torque; is the driving resistance torque equivalent to the driven plate; 、 is the clutch master and driven disc speed; 、 The equivalent moment of inertia of the master and slave ends; 、 is the rotational damping coefficient of the engine and clutch; (2) The Bolza optimization model is solved by pseudospectral method and quadratic programming to obtain the optimal speed change rates of the engine and clutch under three different driving intentions; (3) Offline training is performed using the BP neural network and the rule table. The rate of change signal output by the optimal speed call layer is integrated through the integrator of the preprocessing layer to obtain a reference speed curve for subsequent controller calls. The engine reference speed call rule is as follows: the three parameters of throttle pedal opening, initial acceleration and shift time are used as training feature parameters of the neural network, and the rate of change of the target speed obtained by pseudo-spectral method and quadratic programming is used as label data. When calling, the output of the neural network model is directly integrated to obtain the engine reference speed. The clutch reference speed call rule is as follows: different clutch speed change rate values ​​are given at different throttle openings, and the clutch speed measured by the sensor at the start of the shift is used as the initial value of the reference speed. The reference speed change rate at different throttle openings and the clutch speed at the start of the shift are saved offline in the rule table. When calling, the initial value is integrated to obtain the reference speed of the clutch during the shift process. (4) Using a speed closed-loop scheme, an adaptive sliding mode controller for the engine / clutch during the shifting process is designed. The engine / clutch reference trajectory obtained after integration is used as the target trajectory and input into the shifting model for tracking control; (5) Verification given different driving intentions and environments in Matlab / Simulink.

Images