Dynamic authority distribution system and method for man-machine sharing control

Abstract

The invention provides a man-machine sharing control-oriented dynamic permission allocation system and method. The method comprises the following steps of: establishing a vehicle longitudinal-transverse coupling dynamic model; the robust security constraint is generated through an environment robust control barrier function, and the environment robust control barrier function compensates the influence of sensor measurement uncertainty through a residual error correction term; the performance constraint is generated by controlling a Lyapunov function; based on the robust security constraint and the performance constraint, the change rate of the control permission of the automatic system is used as a core decision variable, an optimization problem is constructed, the robust security constraint is set as a hard constraint, the performance constraint is set as a soft constraint, and the optimization objective is to minimize the permission fluctuation degree and the control error at the same time; and solving the optimization problem to obtain a current control authority weight, and carrying out weighted fusion on the driver control instruction and the automatic system control instruction to generate a final vehicle control instruction.

Description

Technical Field

[0001] This invention belongs to the field of intelligent vehicle control and human-machine hybrid enhanced intelligence technology, specifically relating to a dynamic permission allocation system and method for human-machine shared control. Background Technology

[0002] With the development of intelligent driving technology, vehicle automation has significantly improved road safety and traffic efficiency by reducing human error and optimizing driving trajectories. However, in dynamic and complex open road scenarios, existing systems still face fundamental challenges such as environmental perception uncertainty and real-time decision-making limitations. Against this backdrop, Human-Machine Shared Control (HMSC), which combines human decision-making capabilities with machine control precision, has become a key approach to achieving fully autonomous driving.

[0003] Shared control can be categorized into direct and indirect shared control based on the method of synthesizing the final control input. In direct shared control, human and system inputs directly influence the vehicle's actuators, achieving real-time collaboration. However, direct shared control carries the potential risk of the driver and the automated system vying for control, which could lead to mutual interference and control instability.

[0004] Currently, a key challenge in shared control research lies in the dynamic allocation mechanism of human-machine control permissions. Existing research methods mainly suffer from the following limitations: fixed-weight allocation strategies, while simple to implement, lack clear physical meaning and adaptability; allocation methods based on predefined indicators, while providing some interpretability, heavily rely on the design of difficult-to-optimize metrics; quantitative methods based on driving risk fields attempt to establish a mapping relationship between environmental risk and permission allocation, but are insufficient in multi-source information fusion; and methods that model the shared driving process as a game theory problem face the challenge of computational complexity brought about by non-convex nonlinear optimization. In recent years, allocation methods based on the control Lyapunov function-control barrier function-quadratic programming (CLF-CBF-QP) have demonstrated good safety, interpretability, and real-time performance in typical lane-changing scenarios, providing a new solution for dynamic permission allocation. However, this method still needs improvement in longitudinal-lateral collaborative control and sensor uncertainty handling.

[0005] However, the aforementioned existing technologies have two main limitations: most studies focus only on simplified lateral permission assignment, using steering angle as human input, while ignoring the effects of longitudinal speed variations and longitudinal-lateral coupling; most studies assume ideal perception of the environment, reducing the practical significance of the developed methods. Summary of the Invention

[0006] To address the shortcomings and deficiencies of existing technologies, this invention provides a dynamic permission allocation system and method for human-machine shared vehicles, belonging to the field of intelligent vehicle control technology. This solution aims to solve the technical problems of existing shared control methods that only focus on lateral control, neglecting the impact of sensor uncertainties and unsmooth permission transitions, and constructs a dynamic permission allocation framework that balances lateral and longitudinal collaboration with robust security. Its core innovations are as follows: First, it establishes a longitudinally-laterally coupled nonlinear vehicle dynamics model to uniformly represent the interaction between lateral steering and longitudinal speed changes, providing a precise model foundation for collaborative permission allocation. Second, to address sensor measurement uncertainties, it designs an Environmental Robust Control Barrier Function (ER-CBF), which quantifies sensor error boundaries and introduces residual correction terms to construct robust safety constraints resistant to perception errors. Simultaneously, it combines a control Lyapunov function (CLF) to ensure system control performance and stability. Third, it transforms the permission allocation problem into a convex quadratic programming (QP) optimization problem, using the rate of change of control permissions in the automated system as the core decision variable. Robust safety constraints are set as hard constraints, performance constraints as soft constraints, and relaxation variables are introduced to ensure the solvability of the optimization problem in complex scenarios, achieving a smooth and dynamic transition of permissions. Furthermore, it establishes independent QP optimization problems for lateral and longitudinal control, enabling asynchronous control commands to be responded to, ensuring that at least one dimension of control command is effectively executed at any given time, thus expanding the applicable scenarios of the traditional synchronous collaborative framework. Through joint simulation experiments involving human-in-the-loop and hardware-in-the-loop, this solution significantly improves vehicle handling agility, safety margin, and passenger comfort compared to traditional methods in complex driving scenarios with sensor uncertainties and dynamic speed changes, providing a reliable technical guarantee for the safe and comfortable operation of intelligent vehicles.

[0007] The specific technical solution adopted by this invention to solve its technical problem is as follows:

[0008] A dynamic permission allocation method for human-machine shared control includes:

[0009] Acquire vehicle status information, driver control commands, and automated system control commands;

[0010] A vehicle longitudinal-lateral coupled dynamic model is established to characterize the interaction between the vehicle's lateral and longitudinal motions;

[0011] Based on the coupled dynamics model, robust safety constraints and performance constraints are constructed: the robust safety constraints are generated through an environmental robust control barrier function, which compensates for the influence of sensor measurement uncertainties through a residual correction term; the performance constraints are generated through a control Lyapunov function.

[0012] Based on the robust security constraints and performance constraints, and with the rate of change of control permissions in the automated system as the core decision variable, an optimization problem is constructed. The robust security constraints are set as hard constraints and the performance constraints are set as soft constraints. The optimization objective is to simultaneously minimize the degree of permission fluctuation and control error.

[0013] Solve the optimization problem to obtain the current control authority weight, and then perform weighted fusion of the driver control command and the automation system control command to generate the final vehicle control command.

[0014] Furthermore, the coupled dynamics model is a nonlinear affine dynamics model constructed using a simplified bicycle model, which includes state equations for the vehicle's longitudinal position, lateral position, orientation, and velocity, and considers the coupled influence of sideslip angle and rolling resistance on the vehicle's motion; the sideslip angle is calculated based on the distance from the vehicle's center of gravity to the front axle, the distance from the vehicle's center of gravity to the rear axle, and the steering angle, and the rolling resistance is characterized by a polynomial model containing zero-order terms, first-order velocity terms, and second-order velocity terms.

[0015] Furthermore, the construction process of the environmental robust control barrier function includes: quantifying the uncertainty of the sensor measurement environment state and clarifying its error boundary; calculating the deviation of the safety function and its gradient based on the error boundary; and incorporating the deviation as a residual correction term into the nominal safety constraint to form a robust safety constraint that is resistant to sensor measurement errors.

[0016] Furthermore, the performance objectives of the control Lyapunov function include lateral performance objectives and longitudinal performance objectives; the lateral performance objectives are the squared deviations of the vehicle's lateral position from the reference lateral position and the squared deviations of the vehicle's orientation from the reference orientation; the longitudinal performance objectives are the squared deviations of the vehicle's speed from the reference speed.

[0017] Furthermore, the optimization problem is a convex quadratic programming problem, and the rate of change of the permission is the difference between the control permission weight at the current moment and the control permission weight at the previous moment. The optimization objective is to control the degree of permission fluctuation by minimizing this rate of change. When constructing the convex quadratic programming problem, slack variables are introduced to adjust the soft constraints to ensure the solvability of the optimization problem in complex scenarios.

[0018] Furthermore, separate convex quadratic programming problems are established for vehicle lateral control and longitudinal control, respectively. The control authority weights for lateral and longitudinal control are solved by two independent optimizers. The optimizers respond to control commands that are asynchronous in the lateral and longitudinal dimensions, and ensure that control commands for at least one dimension are effectively executed at any given time.

[0019] Furthermore, the control authority weight ranges from 0 to 1, and the change in the control authority weight is non-negative and does not exceed a preset maximum change.

[0020] Furthermore, based on the control commands fused from the aforementioned permission allocation coefficients, the vehicle is controlled to perform lateral lane change assist and longitudinal adaptive cruise control.

[0021] Furthermore, the slack variable is used to ensure strict compliance with the robust safety constraint by adjusting the degree of satisfaction of the performance constraint when the performance constraint conflicts with the robust safety constraint.

[0022] And, a dynamic permission allocation system for human-machine shared vehicles, comprising:

[0023] The information acquisition module is used to acquire vehicle status information, driver control commands, and automated system control commands;

[0024] The modeling module is used to establish a vehicle longitudinal-lateral coupled dynamics model, which characterizes the interaction between the vehicle's lateral and longitudinal motions;

[0025] The constraint construction module is used to construct robust safety constraints based on the coupled dynamics model by using an environment robust control barrier function and to construct performance constraints by controlling a Lyapunov function. The robust safety constraints compensate for sensor measurement uncertainties through residual correction terms.

[0026] The optimization module is used to construct a convex quadratic programming optimization problem based on the robust security constraints and performance constraints, with the rate of change of control permissions of the automated system as the core decision variable. The robust security constraints are set as hard constraints and the performance constraints are set as soft constraints. The optimization objective is to minimize the degree of permission fluctuation and control error.

[0027] The instruction fusion module is used to solve the convex quadratic programming optimization problem, obtain the control authority weights, and perform weighted fusion of driver control instructions and automated system control instructions to generate the final vehicle control instructions.

[0028] And a computer device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the method described above.

[0029] A non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0030] Compared to existing technologies, this invention and its preferred solution effectively overcome the limitations of traditional human-machine shared control, which only focuses on lateral control and ignores the coupling effect of longitudinal and lateral motion. By constructing a longitudinal-lateral coupled dynamic model, it achieves coordinated management of lateral steering and longitudinal speed control, making permission allocation more aligned with the actual motion patterns of the vehicle and improving adaptability to complex driving scenarios. Addressing the real-world problem of sensor measurement uncertainty, the residual correction mechanism of the environmentally robust control barrier function significantly enhances the system's safety assurance capability in perception-constrained environments and reduces the risk of control instability caused by perception errors. Through convex quadratic programming optimization design with the permission change rate as the core decision variable, a smooth dynamic transition of human-machine control permissions is achieved, avoiding abrupt interference during permission switching and balancing driving stability and passenger comfort. Simultaneously, the design of the independent lateral and longitudinal optimization framework supports efficient collaborative processing of asynchronous control commands, further improving the system's reliability and adaptability in dynamic and complex driving tasks. Overall, this invention, through the organic combination of coupled modeling, robust constraints, and optimized transition mechanisms, not only improves vehicle handling agility and safety margin, but also provides more practical technical support for the safe and comfortable operation of intelligent vehicles. Attached Figure Description

[0031] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0032] Figure 1 This is a simplified bicycle model diagram used in an embodiment of the present invention;

[0033] Figure 2 This is a schematic diagram illustrating the use of three circles surrounding a vehicle to define a safety set in an embodiment of the present invention;

[0034] Figure 3 This is a schematic diagram of the HmIL and HdIL driving simulator platforms used in the embodiments of the present invention;

[0035] Figure 4 This is a theoretical framework diagram of the experimental algorithm in an embodiment of the present invention;

[0036] Figure 5 This is a schematic diagram of a simulation experiment scenario according to an embodiment of the present invention;

[0037] Figure 6 This is a snapshot of the scene in simulation experiment 1 of this invention, where the vehicle in front collides with the vehicle in front at around x=140m;

[0038] Figure 7 This is a comparison diagram of the inputs of the driver and the autonomous driving system in the lateral and longitudinal directions during simulation experiment 1 of this embodiment of the invention;

[0039] Figure 8This is a comparison chart of the method of the present invention and the benchmark method in terms of permission allocation, TTC, and vehicle spacing in simulation experiment 1 of the present invention.

[0040] Figure 9 This is a diagram showing the driver and the input of the autonomous driving system under the robust permission allocation algorithm in simulation experiment 2 of this invention.

[0041] Figure 10 This is a comparison of the minimum lateral TTC of RC-CLF-CBF-QP and C-CLF-CBF-QP in simulation experiment 2 of this embodiment of the invention. Detailed Implementation

[0042] To make the features and advantages of the present invention more apparent and understandable, specific embodiments are described below in detail:

[0043] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0044] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0045] To address the problems of separate lateral and longitudinal control and neglect of sensor uncertainties in existing human-machine shared vehicle control methods, this invention provides a dynamic permission allocation system and method for human-machine shared vehicle control. Specifically, it is a robust dynamic control permission allocation system and method that comprehensively considers lateral and longitudinal coupled dynamics and sensor uncertainties in human-machine shared vehicle control. Addressing the technical limitations of existing control methods that primarily focus on lateral control under a fixed speed assumption and ignore the impact of sensor uncertainties on control decisions, this invention proposes a robust dynamic permission allocation framework capable of simultaneously coordinating and managing lateral and longitudinal control tasks. The system's control modules include a lateral control unit and a longitudinal control unit. The lateral control unit employs a robust environmental control barrier function (ER-CBF) to ensure safety under sensor uncertainty conditions and combines it with a control Lyapunov function (CLF) to achieve lateral control objectives. The longitudinal control unit also uses ER-CBF and CLF to ensure longitudinal motion safety and target tracking performance. Both control units share the same longitudinal-lateral vehicle dynamic model considering dynamic coupling. To solve the smooth transition problem in dynamic permission allocation, the system establishes a convex quadratic programming (QP) optimization model, achieving a natural transition of control permissions by optimizing the permission change rate. This invention also provides a permission allocation method based on this system, including: establishing a longitudinal-lateral coupled dynamic model; designing safety constraints based on ER-CBF and performance constraints based on CLF; constructing a QP optimization problem to solve for the optimal permission allocation strategy; and realizing the collaborative generation of lateral and longitudinal control commands. Hardware-in-the-loop (HIL) and human-in-the-loop (HIL) experiments demonstrate that, in complex scenarios with sensor uncertainties and dynamic speed changes, this invention improves vehicle handling agility and safety compared to traditional fixed permission allocation methods, while significantly enhancing passenger comfort. Compared to existing technologies, this invention breaks through the limitations of traditional single-dimensional control and fixed permission allocation modes. Through coupled dynamic modeling, robust safety constraints, and optimized permission transition mechanisms, it achieves human-machine collaborative control more adapted to real-world driving scenarios, providing technical assurance for the safe and comfortable operation of intelligent vehicles.

[0046] Its implementation can be divided into three parts: longitudinal-lateral coupling dynamic modeling, robust constraint design based on ER-CBF and CLF, and dynamic allocation of permissions based on convex optimization. Its key features are: First, establishing a unified nonlinear dynamic model that simultaneously considers the lateral and longitudinal motion of the vehicle, providing an accurate model foundation for subsequent cooperative control; second, innovatively introducing a robust environmental control barrier function (ER-CBF) to provide strict guarantees for system safety, and combining it with a control Lyapunov function (CLF) to ensure the achievement of control performance objectives; finally, transforming the permission allocation problem into a convex quadratic programming (QP) problem, and achieving smooth and dynamic fusion of driver and automated system control commands by optimizing the rate of change of permission weights. This method overcomes the inherent defects of traditional shared control methods that only focus on lateral control under the assumption of a fixed speed and ignore the impact of sensor uncertainty on control decisions, effectively solving two core problems: longitudinal-lateral control permission cooperative allocation and safety assurance under uncertain environments.

[0047] The overall implementation steps of the present invention are as follows:

[0048] Step 1: Establish a nonlinear affine dynamics model for the vehicle that comprehensively considers longitudinal-lateral coupled dynamics;

[0049] First, a nonlinear affine dynamics model of the vehicle, comprehensively considering longitudinal-lateral coupled dynamics, is established to provide an accurate system description for subsequent access control. Based on this model, lateral and longitudinal automation controllers are designed separately, forming a complete foundation for automation control.

[0050] Step 2: Building a Robust Permission Assignment Framework

[0051] Building upon step one, and addressing the sensor uncertainties present in real-world driving environments, an environmentally robust control barrier function is introduced to construct safety constraints, ensuring system safety under conditions of inaccurate perception. Simultaneously, a control Lyapunov function is employed to guarantee system stability and control performance. These two aspects are integrated into a quadratic programming problem, constructing horizontal and vertical permission allocation optimization problems respectively. The rate of change in control permissions is used as the core decision variable to achieve smooth transition and dynamic allocation of permissions.

[0052] Step 3: System Verification and Performance Evaluation

[0053] Using a human-in-the-loop and hardware-in-the-loop co-simulation experimental platform, a typical urban road driving scenario was constructed to verify the effectiveness of the method. Under different sensor errors and environmental conditions, the system's comprehensive performance in terms of permission allocation smoothness, security assurance, and task completion efficiency was evaluated to ensure the reliability and superiority of the method in practical applications.

[0054] In step one, the ER-CBF is used to construct the safety constraints. The specific implementation method is as follows: First, the uncertainty of the sensor measuring the environmental state is quantitatively analyzed to clarify its error boundary; then, the deviation of the safety function and its gradient is calculated based on the error boundary; finally, the calculated deviation is used as a residual correction term and incorporated into the nominal safety constraint conditions, thereby constructing a robust safety constraint that can withstand the sensor measurement error and ensure the safety of the system in a real perception-limited environment.

[0055] Step two involves constructing a convex quadratic programming problem, which is implemented in two ways: designing the permission optimization objective and setting constraints. In terms of objective design, the objective function of the QP problem aims to simultaneously minimize the fluctuation of control permissions and the tracking deviation of system performance indicators. Differentiated weighting coefficients are used to prioritize ensuring the driver's operational control. Regarding constraint setting, robust safety constraints are set as hard constraints that must be strictly followed, while system performance constraints are set as soft constraints that can be appropriately relaxed. Slack variables are introduced to ensure the solvability of the optimization problem, thereby achieving the optimal balance between safety and performance.

[0056] By establishing independent QP problems for lateral and longitudinal control, the system achieves collaborative processing of asynchronous control commands: the system can respond to time-asynchronous control commands in the lateral and longitudinal dimensions respectively, and dynamically allocate permissions through an independent QP optimizer to ensure that at least one dimension control command is effectively executed at any given time. This extends the traditional synchronous collaborative framework to asynchronous scenarios, significantly enhancing the adaptability and reliability in complex dynamic driving tasks.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] 1. By introducing the Environmental Robust Control Barrier Function (ER-CBF), the uncertainty of sensor measurements is effectively handled, ensuring safe access control and transition in real uncertain environments, and improving the robustness of the system.

[0059] 2. By establishing a shared control model that comprehensively considers longitudinal-lateral coupling dynamics and integrating longitudinal permission allocation into the CLF-CBF-QP framework, a unified processing of longitudinal velocity variation and coupled dynamics is achieved, overcoming the limitation of most existing methods that only focus on lateral control.

[0060] 3. By constructing a convex quadratic programming problem with the rate of change of control permissions as the decision variable, a smooth and dynamic transition of permissions is achieved. While ensuring security and stability, computational efficiency is also taken into account, thus meeting the requirements of real-time control.

[0061] 4. Verification through human-in-the-loop and hardware-in-the-loop joint simulation experiments shows that the method proposed in this invention significantly improves driving agility, safety margin, and passenger comfort.

[0062] This invention further demonstrates the specific implementation of a dynamic permission allocation system for human-machine shared control through embodiments. The system runs on the HmIL and HdIL driving simulator platforms, which consist of PanoSim high-fidelity vehicle simulation software and a Logitech G29 steering wheel. The core algorithm of the permission allocation calculation module executes the following steps:

[0063] Step 1: System Modeling and Controller Design

[0064] First, a nonlinear affine dynamics model of the vehicle is established, comprehensively considering the longitudinal-lateral coupled dynamics. This model fully accounts for the mutual coupling effect between longitudinal and lateral dynamics in actual vehicle motion. Based on this model, lateral and longitudinal automated controllers are designed respectively: lateral steering control uses a model predictive control algorithm to achieve precise lane-changing assistance, ensuring the smoothness and safety of the lane-changing process by predicting the vehicle state over a future period and optimizing the control sequence; longitudinal speed control is based on the CLF-CBF-QP framework to achieve adaptive cruise control, ensuring system stability by constructing a control Lyapunov function and using a control barrier function to ensure a safe distance from the vehicle in front.

[0065] This invention uses a simplified bicycle model of a four-wheeled vehicle to design an automatic controller, where the left and right wheels of the front and rear axles are concentrated on a single wheel, such as... Figure 1 As shown. This simplified bicycle model can be described as follows:

[0066]

[0067] In the Cartesian coordinate system, x and y represent the vehicle's longitudinal and lateral positions, respectively, while ψ represents the vehicle's orientation. v represents the velocity of the vehicle's center of gravity, m is the vehicle's mass, δ is the steering angle, and the angle between v and the vehicle's longitudinal axis is:

[0068]

[0069] Where u is the control input, consisting of two elements: driving force F. u And the steering angle δ, which correspond to the brake pedal and steering gear of the car, respectively. f and l r F describes the distance from the vehicle's center of gravity to the front and rear axles. r (v) is aerodynamic drag, also known as rolling drag, which is approximately:

[0070]

[0071] Model predictive control (MPC) is used in the implementation of ALK, using a bicycle model to predict the future vehicle state ξ. mpc=[x,y,ψ] T And optimize control input u mpc =[δ] T For longitudinal ACC, the goal of the autonomous driving system is to match the speed of the vehicle in front with the speed of the vehicle in front when the distance between the two vehicles is less than the safe distance. This embodiment uses the classic CLF-CBF-QP type ACC algorithm. Considering the nonlinear model, the vehicle state is defined as: Where v represents the vehicle speed, v ref This represents the driver's desired speed or the speed of the vehicle in front. This is a variable that depends on the distance between the two vehicles and the set safe distance. Δx is the distance between the two vehicles. The input is defined as u=[F u ] T This is the driving or braking force applied by the driver.

[0072] Step 2: Building a Robust Permission Assignment Framework

[0073] Control barrier functions have been widely used in critical safety systems, but their control input design relies on measurements without considering uncertainties. In practice, sensor measurements often exhibit uncertainties or inaccuracies; erroneous sensor data can weaken the allocation of drive permissions and even lead to unsafe behavior. Therefore, this paper addresses these issues by designing an Environmentally Robust Control Barrier Function (ER-CBF) framework to compute control inputs.

[0074] Suppose h(x,xs) is the ER-CBF of the system, where xs∈R p This represents the state of the dynamic environment. In this embodiment, the security sets S and h(x,xs) are redefined as follows: ,That This represents a measurement of the state of the environment. This corresponds to the uncertainty in the measurement state. CBF error is quantified by estimating the state, gradient, and time derivative of the surrounding dynamic system. Let:

[0075]

[0076] If there exists an extended class-K function α that is effective for all... Both and are defined and satisfy:

[0077]

[0078] in It is by The residual terms resulting from the differences between them are given as follows:

[0079]

[0080] in, The nominal safe input is obtained by solving the CBF-QP problem without considering environmental uncertainties, while It is the upper limit of the nominal control correction item.

[0081] In lateral and longitudinal coupling control, safety constraints are implemented through the ER-CBF framework to avoid vehicles in adjacent lanes and vehicles ahead in the current lane. For example... Figure 2 As shown, how to construct the distance constraints between the vehicle and the circles enclosing other vehicles to define the safety set can be mathematically represented as follows: Where p = [x, y] T Let i represent the pose of the vehicle, and 1 ≤ i ≤ 1. , 1≤j≤ This indicates that the vehicle was... A circle surrounds the obstacle vehicle. A circle surrounds the vehicle. △r is the lateral safety distance between vehicles, while the longitudinal safety distance is determined by the vehicle's speed and preview (reaction) time. CBF can be designed as follows:

[0082]

[0083] in T is the center of the j-th circle surrounding the obstacle vehicle. h It is the reaction time, V s 'v' is the speed of the vehicle in front, and 'v' is the speed of the ego vehicle.

[0084] In addition to the safety guarantees provided by ER-CBF, this embodiment introduces CLF to select the control input to achieve system stability. Similar to CBF, a positive definite, continuous, and differentiable function is defined. There exists a κ function λ such that:

[0085]

[0086] CLF stands for Meeting Control Performance Objectives. This embodiment considers assisted steering and cruise scenarios under lateral-longitudinal coupling control. The control objective of the automated system is to maintain safe but variable stable lane keeping control, whether before or after possible lane change maneuvers. Incorporating this objective into CLF, we get:

[0087]

[0088] in These are reference values ​​for lateral position, direction, and longitudinal velocity, respectively.

[0089] By integrating the CLF and ER-CBF into a single QP framework, a trade-off between control performance and safety constraints can be achieved. Specifically, the CLF is relaxed to a soft constraint to accommodate flexible tracking of the control objective, while the ER-CBF is formulated as a hard constraint to ensure strict satisfaction of the safety boundary. This embodiment designs a novel rCLF-CBF-QP scheme for shared control of longitudinal and lateral motion. This is achieved by designing two independent QP problems that share the same coupled dynamics. These two QP problems optimize the rate of change of control commands constrained by the CLF and ER-CBF in the longitudinal and lateral dimensions, respectively. The specific formulas are as follows:

[0090]

[0091] Where Δq=q t -q t-1 Solving for q at a fixed sampling frequency. t and q t-1 This refers to the control rights at adjacent time points separated by a time interval of Δt. For simplicity, the time dependency t has been omitted. It is important to note that the primary decision variable of the algorithm is the rate of change of control rights Δq, not u. r This is because the driver inputs u. h and automated control u a It is known in advance. The final control input u r It is a q-weighted combination of the two, which is the essence of shared control. △q1 and △q2 represent the rate of change of driving authority in the lateral and longitudinal directions of the autonomous driving system, respectively. The QP cost function combines q... t and (1-q t ), weighted by Wh and Wa, representing preferences for manual control and automatic control, respectively.

[0092] This embodiment further provides a Figure 3 The experimental environment shown and Figure 4 The theoretical architecture of the experimental algorithm is demonstrated. The driving simulator includes a Logitech G29 steering wheel, a high-resolution screen, and the high-fidelity vehicle simulation software PanoSim. The effectiveness of the method is verified by constructing a typical three-lane urban road driving scenario using a human-in-the-loop and hardware-in-the-loop co-simulation experimental platform, where each lane is 3.2 meters wide. The ego-vehicle is 4.54 meters long and 2.2 meters wide. Two sets of comparative experiments are designed, each targeting one of two benchmark methods, to verify the performance differences between the proposed method and each benchmark method. The two sets of comparative experiments are as follows:

[0093] (1) In an ideal scenario without sensor uncertainty, the proposed method C-CLF-CBF-QP (C stands for Coupled Dynamics) is compared with the CLF-CBF-QP authority allocation method as a baseline method used only for lateral shared control, while longitudinal control employs fixed-weight fusion of driver input, i.e., static allocation. Manual deceleration was excluded in this experiment. For a fair comparison, the proposed method and the baseline method used the same driver input.

[0094] (2) The proposed method is compared with a scenario lacking robustness (i.e., without ER-CBF) in a real-world setting with sensor uncertainty. A scenario with perceptual uncertainty is constructed by adding uniformly distributed random errors over a range of (0,10) cm to the lateral and longitudinal distance sensors. The two methods are compared in terms of safety constraint satisfaction, control command smoothness, and task completion rate.

[0095] Experimental scenarios such as Figure 5 As shown: The driver is traveling at an initial speed of 18 m / s, and a similar vehicle ahead is traveling at a constant speed of 20 m / s; in the adjacent lane, there is a vehicle of the same size but with a different starting position, the same initial speed, and gradually accelerating. The driver attempts to switch to the left lane by accelerating. The vehicle parameters and algorithm hyperparameters used in all scenarios are shown in Table 1.

[0096] Table 1. Vehicle parameters and algorithm parameters in the simulation experiment.

[0097]

[0098] To quantify safety, collision time (TTC) was calculated for all scenarios using the proposed method and baseline methods. In lane-change scenarios, the TTC was calculated as follows:

[0099] Vertical TTC: Horizontal TTC:

[0100] in and represent the longitudinal position and speed of adjacent (obstacle) vehicles, respectively. The same rule applies to the lateral definition.

[0101] In Experiment 1, the driver's objective was to accelerate rapidly, increasing the horizontal distance to the vehicle in the left lane, and then complete a left lane change maneuver. Initially, the driver rapidly accelerated the vehicle; at x=90 meters, the distance to the vehicle in front fell below a safe threshold, prompting a transfer of longitudinal control to the automatic system, which then decelerated to match the speed of the vehicle in front. At x=140 meters, the driver attempted to change lanes to the left lane, but the vehicle in the left lane was rapidly approaching from behind. During the lane change, the vehicle detected that the distance to the vehicle in the left lane was approaching the safe threshold, causing the lane change to fail; the vehicle continued to stay centered in the current lane. Under the same input conditions, the benchmark algorithm exhibited significant performance differences, ultimately leading to a collision between the controlled vehicle and the vehicle in the middle lane at x=140 meters, as shown below. Figure 6 As shown. Figure 7 The results show lateral and longitudinal driver inputs as a function of displacement: longitudinal inputs from human drivers are completely consistent; the baseline algorithm does not include longitudinal inputs from the automatic system. Figure 8 The study further demonstrates the changes in lateral and longitudinal control allocation, TTC, the distance between the vehicle and the vehicle in front, and the lateral distance between the vehicle and the vehicle in the left lane. As a function of displacement, the proposed algorithm's TTC exhibits a trend of "first decreasing and then increasing," while the distance between the main vehicle and the vehicle in front synchronously shows a pattern of "first shortening and then stabilizing." In contrast, for the baseline algorithm, the distance between the two vehicles continuously shortens, and the TTC continuously decreases until a collision occurs.

[0102] Experiment 2 proceeded as follows: In the initial stage, the driver accelerated the vehicle. When the distance reached x=85 meters, the system detected that the distance between the vehicle and the vehicle in front was approaching a safe threshold and gradually transferred driving control to the automatic driving system, subsequently decelerating to match the speed of the vehicle in front. When the distance reached x=145 meters, the driver attempted to change lanes. At this time, a vehicle in the left lane rapidly approached from the left rear. During the lane change, the vehicle detected that the distance to the vehicle on the left was approaching a safe threshold, therefore the lane change failed, and the vehicle continued to travel in the center of the current lane. Figure 9 It displays input data from the driver and the automated system; Figure 10 This includes lateral and longitudinal drive permission allocation, TTC (Traffic Control Center), and distance data to the vehicle in front and vehicles in the left lane.

[0103] Unlike Experiment 1, the results of C-CLF-CBF-QP in Experiment 2 were not found in Figure 9 and Figure 10The difference in lane change timing between the two algorithms is due to variations in longitudinal speed. If the same human input is forced to trigger the lane change, C-CLF-CBF-QP executes the lane change earlier than RC-CLF-CBF-QP. At this point, the distance between the vehicle in the left lane and the driver is still greater than the safety threshold, allowing the lane change to be "successfully" completed. However, this leads to an unfair comparison. Therefore, for Experiment 2, multiple repeated tests were conducted using similar but different human inputs, and the results are shown in Table 2.

[0104] Table 2 Comparison of predicted collision times in Simulation Experiment 2

[0105]

[0106] Based on the same inventive concept, this invention also provides a computer device, comprising: one or more processors, and a memory for storing one or more computer programs; the programs include program instructions, and the processor executes the program instructions stored in the memory. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, used to implement one or more instructions, specifically for loading and executing one or more instructions stored in a computer storage medium to implement the above-described method.

[0107] It should be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium storing a computer program, which, when executed by a processor, performs the above-described method. This storage medium can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0108] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0109] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

[0110] This invention is not limited to the above-described preferred embodiments. Anyone inspired by this invention can derive other forms of dynamic permission allocation systems and methods for human-machine shared control. All equivalent changes and modifications made within the scope of the claims of this invention shall fall within the scope of this invention.

Claims

1. A dynamic permission allocation method for human-machine shared control, characterized in that, include: Acquire vehicle status information, driver control commands, and automated system control commands; A vehicle longitudinal-lateral coupled dynamic model is established to characterize the interaction between the vehicle's lateral and longitudinal motions; Based on the coupled dynamics model, robust safety constraints and performance constraints are constructed: the robust safety constraints are generated through an environmental robust control barrier function, which compensates for the influence of sensor measurement uncertainties through a residual correction term; the performance constraints are generated through a control Lyapunov function. Based on the robust security constraints and performance constraints, and with the rate of change of control permissions in the automated system as the core decision variable, an optimization problem is constructed. The robust security constraints are set as hard constraints and the performance constraints are set as soft constraints. The optimization objective is to simultaneously minimize the degree of permission fluctuation and control error. Solve the optimization problem to obtain the current control authority weight, and then perform weighted fusion of the driver control command and the automation system control command to generate the final vehicle control command.

2. The dynamic permission allocation method for human-machine shared control according to claim 1, characterized in that: The coupled dynamics model is a nonlinear affine dynamics model constructed using a simplified bicycle model. It includes state equations for the vehicle's longitudinal position, lateral position, orientation, and velocity, and considers the coupled influence of sideslip angle and rolling resistance on the vehicle's motion. The sideslip angle is calculated based on the distance from the vehicle's center of gravity to the front axle, the distance from the vehicle's center of gravity to the rear axle, and the steering angle. The rolling resistance is characterized by a polynomial model containing zero-order, first-order, and second-order velocity terms.

3. The dynamic permission allocation method for human-machine shared control according to claim 1, characterized in that: The construction process of the environmental robust control barrier function includes: quantifying the uncertainty of the sensor measurement environment state and clarifying its error boundary; calculating the deviation of the safety function and its gradient based on the error boundary; and incorporating the deviation as a residual correction term into the nominal safety constraint to form a robust safety constraint that is resistant to sensor measurement errors.

4. The dynamic permission allocation method for human-machine shared control according to claim 1, characterized in that: The performance objectives of the control Lyapunov function include lateral performance objectives and longitudinal performance objectives; the lateral performance objectives are the squared deviations of the vehicle's lateral position from the reference lateral position and the squared deviations of the vehicle's orientation from the reference orientation; the longitudinal performance objectives are the squared deviations of the vehicle's speed from the reference speed.

5. A dynamic permission allocation method for human-machine shared control according to claim 1, characterized in that: The optimization problem is a convex quadratic programming problem. The rate of change of the permissions is the difference between the control permission weight at the current time and the control permission weight at the previous time. The optimization objective is to control the degree of permission fluctuation by minimizing this rate of change. When constructing the convex quadratic programming problem, slack variables are introduced to adjust the soft constraints to ensure the solvability of the optimization problem in complex scenarios.

6. A dynamic permission allocation method for human-machine shared control according to claim 5, characterized in that: Independent convex quadratic programming problems are established for vehicle lateral control and longitudinal control, respectively. The control authority weights for lateral and longitudinal control are solved by two independent optimizers. The optimizers respond to control commands that are asynchronous in the lateral and longitudinal dimensions, and ensure that control commands for at least one dimension are effectively executed at any given time.

7. A dynamic permission allocation method for human-machine shared control according to claim 1, characterized in that: The control authority weight ranges from 0 to 1, and the change in the control authority weight is non-negative and does not exceed the preset maximum change.

8. A dynamic permission allocation method for human-machine shared control according to claim 2, characterized in that: Based on the control commands fused from the aforementioned permission allocation coefficients, the vehicle is controlled to perform lateral lane change assist and longitudinal adaptive cruise control.

9. A dynamic permission allocation method for human-machine shared control according to claim 5, characterized in that: The slack variable is used to ensure strict compliance with the robust safety constraint by adjusting the degree to which the performance constraint is satisfied when the performance constraint conflicts with the robust safety constraint.

10. A dynamic permission allocation system for human-machine shared vehicles, characterized in that, include: The information acquisition module is used to acquire vehicle status information, driver control commands, and automated system control commands; The modeling module is used to establish a vehicle longitudinal-lateral coupled dynamics model, which characterizes the interaction between the vehicle's lateral and longitudinal motions; The constraint construction module is used to construct robust safety constraints based on the coupled dynamics model by using an environment robust control barrier function and to construct performance constraints by controlling a Lyapunov function. The robust safety constraints compensate for sensor measurement uncertainties through residual correction terms. The optimization module is used to construct a convex quadratic programming optimization problem based on the robust security constraints and performance constraints, with the rate of change of control permissions of the automated system as the core decision variable. The robust security constraints are set as hard constraints and the performance constraints are set as soft constraints. The optimization objective is to minimize the degree of permission fluctuation and control error. The instruction fusion module is used to solve the convex quadratic programming optimization problem, obtain the control authority weights, and perform weighted fusion of driver control instructions and automated system control instructions to generate the final vehicle control instructions.

Images