CN-122009148-A - Unmanned longitudinal control method of concrete transport vehicle

CN122009148ACN 122009148 ACN122009148 ACN 122009148ACN-122009148-A

Abstract

The invention relates to the technical field of unmanned longitudinal control, in particular to an unmanned longitudinal control method of a concrete truck, which comprises the steps of acquiring vehicle running state data information of the concrete truck in real time; the method comprises the steps of inputting vehicle running state data information into a pre-trained longitudinal speed control strategy network, obtaining a function attenuation rate after the pre-trained longitudinal speed control strategy network is designed for an auxiliary sliding mode controller according to a longitudinal dynamics model of the whole vehicle, constructing a mixed integer model predictive controller based on the auxiliary sliding mode controller to optimize driving torque and braking torque, solving the mixed integer model predictive controller according to a near-end optimization strategy of a discrete and continuous mixed action space, and sending a driving control instruction to an executing mechanism of a concrete transport vehicle. The unmanned longitudinal control method of the concrete transport vehicle can improve the robustness, the instantaneity and the control precision of the automatic driving of the concrete transport vehicle.

Inventors

LUO CHUJUN
SUN HAIMING
XIE GUOTAO
ZHANG XIAOGUANG
YANG ZEYU
LAN GUOYUAN
CHANG YUAN
MA YAN
HUANG TAO

Assignees

中国葛洲坝集团股份有限公司
湖南大学无锡智能控制研究院

Dates

Publication Date: 20260512
Application Date: 20260313

Claims (10)

1. An unmanned longitudinal control method of a concrete truck, comprising: acquiring vehicle running state data information of a concrete transport vehicle in real time, wherein the vehicle running state data information at least comprises vehicle speed, driving torque, braking torque and inclination sensor data; Inputting the vehicle running state data information into a pre-trained longitudinal speed control strategy network to obtain a driving control instruction, wherein the pre-trained longitudinal speed control strategy network is obtained by designing an auxiliary sliding mode controller according to a longitudinal dynamics model of the whole vehicle to obtain a function attenuation rate for stability constraint, constructing a mixed integer model prediction controller based on the auxiliary sliding mode controller to optimize driving torque and braking torque, and solving the mixed integer model prediction controller according to a discrete and continuous mixed action space near-end optimization strategy; The driving control instruction is sent to an executing mechanism of the concrete transport vehicle so as to realize unmanned longitudinal control of the concrete transport vehicle; and updating the pre-trained longitudinal speed control strategy network according to the vehicle control response result of the concrete transport vehicle and the driving control instruction.
2. The unmanned longitudinal control method of a concrete truck according to claim 1, wherein the pre-trained longitudinal speed control strategy network is obtained by designing an auxiliary sliding mode controller according to a longitudinal dynamics model of the whole truck to obtain a function attenuation rate for stability constraint, constructing a mixed integer model predictive controller based on the auxiliary sliding mode controller to optimize driving torque and braking torque, and solving the mixed integer model predictive controller according to a discrete and continuous mixed motion space near-end optimization strategy, and comprises the following steps: Constructing a dynamic model of the whole longitudinal movement and an actuator layer of the concrete transportation vehicle according to the whole longitudinal dynamic model of the concrete transportation vehicle and the rolling dynamic model of the front wheels and the rear wheels of the concrete transportation vehicle; respectively constructing a sliding mode controller under a driving working condition and a sliding mode controller under a braking working condition according to the longitudinal movement of the whole vehicle of the concrete transport vehicle and a dynamic model of an actuator layer, and respectively obtaining a function attenuation rate for stability constraint under the driving working condition and a function attenuation rate for stability constraint under the braking working condition; constructing a mixed integer model predictive controller according to the sliding mode controller under the driving working condition and the sliding mode controller under the braking working condition; And performing offline training on the mixed integer model predictive controller according to a near-end strategy optimization algorithm of an offline continuous mixed action space to obtain a pre-trained longitudinal speed control strategy network.
3. The unmanned longitudinal control method of a concrete transporting vehicle according to claim 2, wherein constructing the dynamic model of the whole longitudinal movement and the actuator layer of the concrete transporting vehicle according to the whole longitudinal dynamic model of the concrete transporting vehicle and the rolling dynamic model of the front wheels and the rear wheels of the concrete transporting vehicle comprises: Constructing a whole vehicle longitudinal dynamics model of the concrete transportation vehicle; Respectively constructing a front wheel rolling dynamics model and a rear wheel rolling dynamics model of the concrete transportation vehicle; determining the relation between the engine speed and the speed of the vehicle according to the transportation scene of the concrete transportation vehicle; Determining a hysteresis coefficient under a driving working condition and a hysteresis coefficient under a braking working condition; And obtaining the dynamic model of the whole vehicle longitudinal movement and the actuator layer of the concrete transportation vehicle according to the whole vehicle longitudinal dynamic model, the front wheel rolling dynamic model, the rear wheel rolling dynamic model, the relation between the rotating speed and the speed of the vehicle engine, the hysteresis coefficient under the driving working condition and the hysteresis coefficient under the braking working condition of the concrete transportation vehicle.
4. The unmanned longitudinal control method of a concrete transporting vehicle according to claim 2, wherein the constructing a slip-form controller under driving conditions and a slip-form controller under braking conditions according to the longitudinal movement of the whole vehicle of the concrete transporting vehicle and the dynamic model of the actuator layer, respectively, and obtaining the function attenuation rate for stability constraint under driving conditions and the function attenuation rate for stability constraint under braking conditions, respectively, comprises: respectively determining sliding mode variables under a driving working condition and a braking working condition according to the longitudinal movement of the whole vehicle of the concrete transport vehicle and a dynamic model of an actuator layer; determining a sliding mode control law under a driving working condition according to a sliding mode variable under the driving working condition, and determining the sliding mode control law under a braking working condition according to the sliding mode variable under the braking working condition; and obtaining the function attenuation rate for stability constraint under the driving working condition according to the sliding mode control law under the driving working condition, and obtaining the function attenuation rate for stability constraint under the braking working condition according to the sliding mode control law under the braking working condition.
5. The unmanned longitudinal control method of a concrete transporting vehicle according to claim 4, wherein determining slip form variables under driving conditions and braking conditions according to a whole vehicle longitudinal motion of the concrete transporting vehicle and a dynamic model of an actuator layer, respectively, comprises: According to the longitudinal movement of the whole vehicle of the concrete transportation vehicle and the dynamic model of an actuator layer, respectively determining an adaptive update law under a driving working condition and an adaptive update law under a braking working condition, wherein the expression of the adaptive update law under the driving working condition is as follows: , Wherein, the Representing an adaptive update law under driving conditions, Representing the anti-control input saturation variable under driving conditions, Representing the update coefficient under the driving condition, Representing the engine-to-wheel transmission ratio, Representing the dynamics coefficients in the dynamics model of the whole car longitudinal movement and the actuator layer of the concrete transportation vehicle, Represents the driving hysteresis coefficient in the dynamic model of the whole vehicle longitudinal movement and the actuator layer of the concrete transportation vehicle, Which indicates an engine torque control command, Representing a control input saturation function; The expression of the self-adaptive update law under the braking working condition is as follows: , Wherein, the Indicating an adaptive update law under braking conditions, Representing the anti-control input saturation variable under braking conditions, Representing the updated coefficients under the braking conditions, Represents the braking hysteresis coefficient in the dynamic model of the whole vehicle longitudinal movement and the actuator layer of the concrete transportation vehicle, Representing a brake system torque control command; According to the self-adaptive updating law under the driving working condition and the self-adaptive updating law under the braking working condition, respectively designing a sliding mode surface under the driving working condition and a sliding mode surface under the braking working condition, and obtaining a sliding mode variable under the driving working condition and a sliding mode variable under the braking working condition, wherein the sliding mode variable under the driving working condition is expressed as: , Wherein, the Represents the sliding mode variable under the driving working condition, The coefficient of the sliding mode surface is represented, The control error is indicated as such, A first derivative representing a control error; The sliding mode variable under the braking working condition is expressed as: , Wherein, the And represents the slip form variable under the braking condition.
6. The unmanned longitudinal control method of a concrete truck of claim 4, wherein determining the slip form control law under driving conditions from slip form variables under driving conditions and determining the slip form control law under braking conditions from slip form variables under braking conditions comprises: Determining a sliding mode approach law under a driving working condition according to a sliding mode variable under the driving working condition, wherein the sliding mode approach law under the driving working condition has the following expression: , Wherein, the Represents the approach law of the sliding mode under the driving working condition, The parameters of the sliding mode surface are represented, The sign function is represented by a sign function, Representing the upper bound of variation of the first derivative of the system uncertainty; Determining a sliding mode control law under a driving working condition according to the sliding mode approach law under the driving working condition, wherein the sliding mode control law under the driving working condition has the following expression: , Wherein, the Which indicates an engine torque control command, Representing the actual torque of the engine, Representing the second derivative of the desired speed, A first derivative representing a coefficient of resistance to vehicle travel; Determining a sliding mode approach law under a braking working condition according to a sliding mode variable under the braking working condition, wherein the sliding mode approach law under the braking working condition has the following expression: , Wherein, the The sliding mode approach law under the braking working condition is shown; determining a sliding mode control law under a braking working condition according to the sliding mode approach law under the braking working condition, wherein the sliding mode control law under the braking working condition has the following expression: , Wherein, the Indicating a braking system torque control command, Indicating the total braking torque.
7. The unmanned longitudinal control method of a concrete transporter according to claim 2, wherein constructing the hybrid integer model predictive controller from the slip-form controller under driving conditions and the slip-form controller under braking conditions comprises: Determining a cost function of the mixed integer model predictive controller according to a control target of the concrete transportation vehicle; And determining a constraint target of the mixed integer model predictive controller according to the longitudinal dynamics constraint of the vehicle, the hysteresis constraint of the driving system, the hysteresis constraint of the braking system, the function attenuation rate constraint and the control input saturation constraint.
8. The unmanned longitudinal control method of a concrete transporter according to claim 2, wherein the offline training of the mixed integer model predictive controller according to a near-end policy optimization algorithm of an offline continuous mixed motion space, to obtain a pre-trained longitudinal speed control policy network, comprises: Constructing a mixed motion space vector according to the speed error and the derivative thereof, the anti-saturation state variable of the driving system, the anti-saturation state variable of the braking system, the decision mode at the last moment and the current road gradient, and defining the motion of the mixed motion space vector to comprise discrete motion and continuous motion; Constructing a strategy network and a value network, wherein the strategy network comprises a discrete strategy header and a continuous strategy header; Correcting probability distribution of the strategy network according to an action mask mechanism based on function attenuation constraint so that a moment instruction obtained by sampling is positioned in a stability area; determining a reward function; Training the strategy network and the value network according to the reward function to construct a hybrid action space loss function, and obtaining a pre-trained longitudinal speed control strategy network.
9. The unmanned vertical control method of a concrete transporter of claim 8, wherein training the strategy network and the value network according to the reward function to construct the hybrid action space loss function comprises: Initializing a strategy network and a value network, and determining physical parameters and sliding film surface coefficients of a concrete transportation vehicle; performing track sampling and performing stability constraint embedding; calculating a dominance value and a target value at each moment according to the generalized dominance estimation; calculating a mixed space probability ratio, constructing a mixed action space loss function, and simultaneously updating strategy network parameters and value network parameters through back propagation; repeating the steps until the strategy network converges.
10. The unmanned longitudinal control method of a concrete transporter according to claim 9, wherein performing track sampling and stability constraint embedding comprises: acquiring a mixed motion space vector at the current moment; Determining a stability boundary of a function attenuation constraint; determining an output mode of the policy network according to the action mask; the motion at the current time is applied to the vehicle model and the hybrid motion space vector at the next time is updated according to the kinetic equation, and the current reward is calculated according to the reward function.

Description

Unmanned longitudinal control method of concrete transport vehicle Technical Field The invention relates to the technical field of unmanned longitudinal control, in particular to an unmanned longitudinal control method of a concrete transport vehicle. Background Longitudinal motion control of a concrete truck is a fundamental and important task in an unmanned system, and aims to drive the vehicle to accurately track a preset speed through coordinated control of the opening degree of an engine throttle valve and the braking pressure. For unmanned concrete trucks operating in complex environments such as mines, construction sites, etc., longitudinal control is challenging. Firstly, the concrete transport vehicle has obvious dynamics nonlinearity, and the engine driving system and the braking system have obvious execution response lag, secondly, the inertia parameter of the whole vehicle has extremely strong uncertainty due to the dynamic change of the concrete load (such as no load, full load and mass fluctuation in the unloading process) in the transport process, and thirdly, the vehicle is influenced by the coupling of road gradient fluctuation, rolling resistance and nonlinear air resistance, so that the system is continuously and externally disturbed. The current control schemes can be categorized into two types, one is a scheme based on an analytic control law (such as PID control, sliding mode control, etc.), and the other is a scheme based on optimal control (such as model predictive control). Methods based on analytic control laws often employ explicit expressions as control quantities, possibly including feedback terms, feedforward terms, or more complex algebraic terms. Some analytic control methods based on analytic control laws in the prior art can give out explicit control laws, ensure control instantaneity and have good control effect under specific working conditions, but still have certain limitations that 1) because a vehicle is generally not allowed to drive and brake simultaneously, the control schemes based on analytic control laws all need to define switching logic of driving and braking additionally so as to judge whether to carry out driving torque control or braking torque control. However, in a concrete transportation scenario, a fixed switching threshold may be detrimental to reasonably switching modes under different road grade, different loading mass conditions. The working condition-adaptive switching threshold is formulated, and the working condition-adaptive switching threshold is complex in engineering application. 2) Drive and brake are often two sets of control schemes, requiring different control parameters. The control parameters directly influence the control precision and response speed, and the setting parameters are required to be independently optimized in actual engineering application, so that the flow is relatively complicated. 3) There are physical limits to both the drive torque and the brake torque of a concrete-transporting vehicle. The above control strategy does not consider input saturation when analyzing the stability of the closed loop system, and mostly adopts a direct cut-off post-processing mode, so that the cut-off control input can not be ensured to meet the closed loop stability in theory, and adverse phenomena such as control oscillation and instability can be caused. The optimal control-based method often models the driving and braking control problem as a whole optimization problem, solves the optimal driving and braking control quantity at the same time, and introduces an integer decision variable to decide whether the driving mode or the braking mode is currently adopted. Although the scheme based on optimal control in the prior art can realize self-optimizing of driving and braking torque, compared with an analytic control law, the scheme can improve control efficiency, the scheme still has the following limitations that 1) the environment is worse in a concrete transportation scene, the road gradient and the vehicle loading quality change in real time, and model mismatch can exist in longitudinal dynamics of the vehicle. Aiming at the uncertainty factors, the current optimal control strategy often ignores the stability of the closed-loop system under the uncertainty disturbance, lacks robustness guarantee, and is difficult to adapt to the severe environment and the complex working condition of the concrete transportation scene. 2) The feasibility of the optimization problem proves that the feasibility of the initial time is often dependent on the setting of the prediction time domain. With the increase of the prediction time domain, the calculation load is also increased, and the real-time requirement of engineering application is difficult to meet. 3) Some studies have simplified the problem by simplifying the dynamic modeling or linearization, but the decrease in model accuracy will inevitably decrease control accuracy. Therefore, how to solv