CN118011826A - UAV control strategy safety verification and correction method and system - Google Patents

Abstract

The present invention relates to a method and system for verifying and correcting the safety of a drone control strategy, wherein the method comprises: obtaining the current drone system status information, the unverified potential unsafe actions output by the original drone flight control strategy of the drone system, and the safety constraint boundary of the drone adaptive adjustment; based on the optimization target norm, establishing and solving a finite time domain optimization problem with drone dynamics and safety constraints, and obtaining safety control actions for the drone system flight control; based on the safety constraint boundary of the drone adaptive adjustment, establishing and updating the safety constraints that the drone state needs to meet; iteratively calculating within a finite time domain, and obtaining the safety state sequence of the drone within the future finite time domain and the corresponding safety control action sequence. The present invention verifies and corrects the potential unsafe drone flight control strategy, and achieves the purpose of safe and maneuverable flight control of the drone system.

Description

UAV control strategy safety verification and correction method and system

Technical Field

The present invention belongs to the technical field of unmanned aerial vehicles, and in particular relates to a method and system for verifying and correcting the safety of unmanned aerial vehicle control strategies.

Background technique

Due to their advantages such as flexibility, agility, and vertical take-off and landing, drones are widely used in industrial and agricultural fields such as search and rescue, circuit inspection, express delivery, aerial photography and mapping, and agricultural and forestry plant protection. Nevertheless, as an aerial unmanned system that is extremely sensitive to dangerous factors such as its own dynamic changes and external environmental disturbances, with the diversification of application scenarios and the increase in environmental complexity, higher requirements and challenges are placed on the safety of drone flight control strategies. However, since safety requirements have increased the conservatism of drone flight control strategies, the trade-off between safety and maneuverability is also a key factor affecting the efficient execution of complex flight missions by drones. For this reason, a safer and more maneuverable drone flight control strategy has become an urgent need for current drone systems. At present, research on the safety of drone flight control strategies is mainly divided into the following methods:

(1) Model-based optimal control method

This type of method usually models the UAV system as a fixed rigid body. On this basis, by abstracting the UAV control strategy into an optimal control problem that satisfies dynamic constraints and various safety constraints, a UAV flight control strategy with safety guarantees is obtained to achieve the purpose of safe flight control of the UAV. By solving a finite time domain model predictive control problem with constraints, a UAV control strategy that can simultaneously satisfy all constraints is obtained, solving the safety challenges encountered by the UAV during flight. However, this method needs to rely on strict and complex mathematical forms to construct the optimization objectives and constraints of the UAV control strategy, which leads to its complex nonlinear structure and is often difficult to obtain its feasible solution through approximate solutions. In addition, the satisfaction of the complex constraints of the UAV control strategy will continuously increase the conservatism of its output control actions, making it unsuitable for the UAV system's demand for flight maneuverability.

(2) Model-free secure reinforcement learning method

As a new type of intelligent self-learning sequential decision-making method, reinforcement learning method has gradually been applied in the field of UAV flight control because it does not rely on complex system models and only interacts with the environment through trial and error. Based on its powerful strategy search and optimization capabilities, it has been gradually applied in the field of UAV flight control. However, most model-free reinforcement learning methods cannot explicitly express the safety constraints that the system needs to meet, so the learned UAV control strategies do not have safety guarantees. Although a class of safety reinforcement learning methods can be constructed using constrained Markov decision processes ^[7] to train UAV control strategies that meet safety requirements, such methods will affect the efficiency of strategy exploration and optimization to a certain extent due to the limitations of safety constraints. In addition, the trained UAV safety control strategies are deeply coupled with the application scenarios and specific flight missions, and lack universality and generalization capabilities for diverse application scenarios. In addition, during the training process of the UAV control strategy, due to the continuous trial and error and exploration of its interaction with the environment, it is impossible to avoid violating safety constraints, and thus its safety during the training process cannot be guaranteed.

(3) Methods based on security verification

Safety verification is another method that verifies the safety of drone control strategies by judging whether they violate safety constraints. At the same time, the safety of drone control strategies is corrected based on the judgment results, thereby achieving safe flight control of drone systems. A recovery strategy that replaces unsafe control strategies is trained through learning methods as a backup strategy in the event of violations of safety constraints, which improves the safety of the control strategy to a certain extent. However, since the recovery strategy is a data-driven learning method, it still has safety risks and cannot provide strict safety guarantees. In order to avoid the safety risks of learning methods, the control strategy is restricted to a safe set with a strict mathematical form by constructing a barrier function. However, due to the difficulty of barrier function construction and high computational complexity, this method cannot meet the real-time requirements of drone control. Some studies have also proposed a safety filter to predict and correct the original control strategy of the system to achieve the purpose of safety control, but since it does not consider the adaptability of the filter to complex and diverse application scenarios, it cannot be applied to drone systems to perform complex flight missions in changing scenarios.

Summary of the invention

The present invention provides a method and system for verifying and correcting the safety of an unmanned aerial vehicle control strategy, aiming to solve at least one of the technical problems existing in the prior art.

The technical solution of the present invention relates to a method and system for verifying and correcting the safety of a drone control strategy. The method for verifying and correcting the safety of a drone control strategy is applied to a drone system, wherein the drone system includes an existing drone flight control strategy. The method comprises the following steps:

S100, obtaining the current UAV system status information, the unverified potential unsafe actions output by the original UAV flight control strategy of the UAV system, and the safety constraint boundary of the UAV adaptive adjustment;

S200, based on the optimization target norm, establish and solve a finite time domain optimization problem with UAV dynamics and safety constraints, and obtain a safety control action for UAV system flight control;

S300, establishing and updating the safety constraints that the drone state needs to satisfy based on the safety constraint boundary of the drone adaptive adjustment;

S400, iteratively calculating step S200 and step S300 within a limited time domain to obtain a safety state sequence and a corresponding safety control action sequence of the UAV within a future limited time domain.

Further, the step S200 includes:

S210, defining an optimization target norm distance of the UAV system based on maximizing the performance of the original UAV flight control strategy while ensuring the safety of its output control action;

S220. Based on the received UAV state information and the predicted safety control action, the UAV dynamics transfer is performed to establish the UAV dynamics constraints of the safety control action within the limited time domain [0, M-1].

Further, in step S210, the optimization target norm distance is:

in, is the unverified potentially unsafe control action output by the original UAV flight control strategy at time t, /> It is the safety control action of the UAV at time t obtained after safety verification and optimization of the correction method.

Furthermore, the original UAV flight control strategy outputs unverified potentially unsafe control actions for:

Among them, _Ft is the combined thrust generated by all the onboard motors in the UAV, and _ωt is the three-axis angular velocity of the UAV in the body coordinate system.

Furthermore, in step S220, for all integers m within a limited time domain, the UAV dynamics constraint is:

in, And m is an integer, s _t+m is the state of the drone at time t+m, /> is the safety control action of the UAV at time t+m, and s _t+m+1 is the state at the next time t+m+1 calculated by the UAV dynamics model f(·).

Further, in step S220, the drone state s _t at time t includes:

s _t =[p _t ,q _t ,v _t ],

Among them, p _t is the position information of the UAV at time t, q _t is the attitude quaternion information of the UAV at time t, and v _t is the linear velocity information of the UAV at time t.

Furthermore, in step S300, the safety constraint of the drone state includes that the drone state within a limited time domain must be within the safety state set S.

The security state set S is:

Among them, s _t is the state of the UAV at time t, B is the weight matrix of the state safety constraint on each state component, is the safety state boundary value that the drone state needs to meet at time t,

in, is the minimum value of the drone's position information at time t, /> is the minimum value of the UAV's attitude quaternion information at time t,/> is the minimum value of the UAV's linear velocity information at time t, /> is the maximum value of the drone's position information at time t, /> is the maximum value of the UAV's attitude quaternion information at time t,/> is the maximum value of the UAV’s linear velocity information at time t.

Furthermore, in step S300, the safety constraints of the drone control actions include that the drone control actions within a limited time domain must be within the safety action set U.

The security action set U includes:

Among them, U represents the set of safety constraints that the drone control action needs to meet,

Where A represents the weight matrix of the control action safety constraint on each control component, is the safety boundary value that the drone control action needs to meet at time t,

in, is the minimum value of the combined thrust generated by all onboard motors in the drone,/> is the maximum value of the combined thrust generated by all the onboard motors in the drone,/> is the minimum value of the three-axis angular velocity of the drone in the body coordinate system, /> It is the maximum value of the three-axis angular velocity of the UAV in the body coordinate system.

Furthermore, at the end time M of the finite time domain, the drone state s _t+M needs to satisfy the state termination set S _end , that is,

s _t+M ∈S _end ,

Among them, S _end is the state termination set, and the state termination set S _end is a subset of the state safety constraint set S, that is,

Among them, S is the set of state safety constraints.

Furthermore, the present invention also proposes a UAV control strategy safety verification and correction system for implementing a UAV control strategy safety verification and correction method, wherein the UAV control strategy safety verification and correction system comprises:

An unmanned aerial vehicle system, the unmanned aerial vehicle system including an existing unmanned aerial vehicle flight control strategy, the existing unmanned aerial vehicle flight control strategy generating unverified potentially unsafe control actions based on unmanned aerial vehicle state information of the unmanned aerial vehicle system to control the flight of the unmanned aerial vehicle system;

A safety verification and correction device, wherein the safety verification and correction device obtains potential unsafe control actions from the original UAV flight control strategy of the UAV system, generates safe control actions and transmits them to the UAV system, and the safety verification and correction device is connected to the UAV system.

Compared with the prior art, the present invention has the following characteristics:

The present invention proposes a unified and universal plug-and-play UAV flight control strategy safety verification and correction method. Based on the idea of optimization theory, the potentially unsafe UAV flight control strategy is verified and corrected by decoupling, aiming to maximize the retention of the original UAV flight control strategy maneuverability, ensure its safety, and achieve the purpose of safe and maneuverable flight control of the UAV system. In addition, for the proposed UAV flight control strategy safety verification and correction method, the present invention designs a corresponding UAV system to deploy and verify its feasibility and effectiveness, and ultimately realizes its application in a variety of real scenarios.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a flow chart of the UAV control strategy safety verification and correction method.

Figure 2 is a flow chart of the establishment of the UAV control strategy safety verification and correction method and the solution of the finite time domain optimization problem with UAV dynamics and safety constraints.

FIG3 is a schematic diagram of the safety verification and correction method and system framework of the UAV flight control strategy.

FIG4 is a schematic diagram of the UAV system architecture of the UAV control strategy safety verification and correction method.

FIG5 is a schematic diagram of the original UAV flight control strategy structure of the UAV control strategy safety verification and correction method.

FIG6 is a schematic diagram of a safety verification and correction method framework for a UAV control strategy safety verification and correction method.

FIG7 is a flowchart of solving the safety verification and correction optimization problem of the UAV control strategy safety verification and correction method.

FIG8 is a schematic diagram of a real-time display interface of UAV control action data during the simulation process of the UAV control strategy safety verification and correction method.

FIG9 is a schematic diagram of the simulation results of the drone takeoff test of the drone control strategy safety verification and correction system in a hall environment with limited height (maximum height 3m).

Figure 10 shows the safety verification and correction system of the UAV control strategy under the condition of limited width (width range [-

Schematic diagram of the simulation results of the drone crossing test in a corridor environment with a diameter of 0.5, 0.5]m.

Figures 11a to 11b are schematic diagrams of the simulation results of the drone tracking trajectory test of the drone control strategy safety verification and correction system in a room environment with limited space (length [-1,1]m, width [-1,1]m, height [0,2]m).

Detailed ways

In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

The concept, specific structure and technical effects of the present invention will be clearly and completely described below in combination with the embodiments and drawings to fully understand the purpose, scheme and effect of the present invention.

It should be noted that, unless otherwise specified, when a feature is referred to as being "fixed" or "connected" to another feature, it may be directly fixed or connected to another feature, or it may be indirectly fixed or connected to another feature. The singular forms "a", "said" and "the" used herein are also intended to include the plural forms, unless the context clearly indicates otherwise. In addition, unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art. The terms used in this specification are intended only to describe specific embodiments and are not intended to limit the invention. The term "and/or" used herein includes any combination of one or more related listed items.

It should be understood that although the terms first, second, third, etc. may be used to describe various elements in the present disclosure, these elements should not be limited to these terms. These terms are only used to distinguish elements of the same type from each other. For example, without departing from the scope of the present disclosure, the first element may also be referred to as the second element, and similarly, the second element may also be referred to as the first element. The use of any and all examples or exemplary language ("for example", "such as", etc.) provided herein is intended only to better illustrate embodiments of the present invention, and unless otherwise required, will not impose limitations on the scope of the present invention. In addition, the industry term "posture" used herein refers to the position and posture of a certain element relative to a spatial coordinate system.

1 to 11 b, an embodiment of the present invention provides a method and system for verifying and correcting the safety of a drone control strategy. The method is applied to a drone system, wherein the drone system includes an existing drone flight control strategy. Referring to FIG. 1 , the method includes the following steps:

S100, obtaining the current UAV system status information, the unverified potential unsafe actions output by the original UAV flight control strategy of the UAV system, and the safety constraint boundary of the UAV adaptive adjustment;

S200, based on the optimization target norm, establish and solve a finite time domain optimization problem with UAV dynamics and safety constraints, and obtain a safety control action for UAV system flight control;

S300, establishing and updating the safety constraints that the drone state needs to satisfy based on the safety constraint boundary of the drone adaptive adjustment;

S400, iteratively calculating step S200 and step S300 within a limited time domain to obtain a safety state sequence and a corresponding safety control action sequence of the UAV within a future limited time domain.

Compared with the prior art, the present invention has the following characteristics:

The present invention proposes a unified and universal plug-and-play UAV flight control strategy safety verification and correction method. Based on the idea of optimization theory, the potentially unsafe UAV flight control strategy is verified and corrected by decoupling, aiming to maximize the retention of the original UAV flight control strategy maneuverability, ensure its safety, and achieve the purpose of safe and maneuverable flight control of the UAV system. In addition, for the proposed UAV flight control strategy safety verification and correction method, the present invention designs a corresponding UAV system to deploy and verify its feasibility and effectiveness, and ultimately realizes its application in a variety of real scenarios.

In view of the defects of existing UAV safe flight control methods, such as high difficulty in solving, strong conservatism, serious coupling, low exploration efficiency, poor versatility, and inability to adapt to diverse and complex scenarios, the main purpose of the present invention is to propose a unified and universal plug-and-play UAV flight control strategy safety verification and correction method, based on the idea of optimization theory, and on this basis, verify and correct the potential unsafe UAV flight control strategy by decoupling, aiming to maximize the retention of the original UAV flight control strategy maneuverability while ensuring its safety, so as to achieve the purpose of safe and maneuverable flight control of the UAV system. In addition, for the proposed UAV flight control strategy safety verification and correction method, the present invention designs a corresponding UAV system to deploy and verify its feasibility and effectiveness, and finally realizes its application in diverse real scenarios.

The technical solution of the present invention brings the following beneficial effects:

(1) Compared with the model-based optimal control method, the present invention does not need to construct complex nonlinear optimization objectives to achieve the flight control purpose of the UAV, thereby greatly reducing the difficulty of solving the optimization problem. At the same time, the present invention mainly emphasizes the safety verification and correction of the potential unsafe UAV flight control strategy on the basis of maximizing the performance of the original UAV flight control strategy. Therefore, the proposed safety verification and correction method is effectively avoided. The impact on the performance of the original UAV flight control strategy, thereby reducing the overall conservatism of the UAV safe flight control strategy.

(2) Compared with the model-free safety reinforcement learning method, the present invention separates the safety constraints that the UAV flight control strategy needs to meet from the training of the safety reinforcement learning control strategy by decoupling it, while avoiding the negative impact of the safety constraints on the strategy search efficiency and the serious coupling of the strategy with specific scenarios and tasks. As a result, the proposed UAV control strategy safety verification and correction method has a unified and universal plug-and-play feature, which can be applied to the safety verification of all potentially unsafe UAV flight control strategies, and ultimately provide safe flight guarantees for the UAV system.

(3) Compared with the existing safety verification-based methods, the present invention is based on a designed UAV system corresponding to the proposed safety verification and correction method, and utilizes the environmental feature perception module in the UAV system to realize adaptive adjustment of the safety constraint boundary, thereby enabling the proposed UAV control strategy safety verification and correction method to have the ability to adapt to diverse and complex environments, greatly expanding its application scenarios.

Specifically, after step S100, after obtaining the unverified potential unsafe control action output by the original UAV flight control strategy, in order to verify and correct its safety, the present invention proposes a safety verification and correction method as shown in Figure 4, which transforms the problem into a finite time domain optimization problem with UAV dynamics and safety constraints through the idea of optimization theory, and finally outputs the UAV control action with safety guarantee. The finite time domain optimization problem mainly consists of three parts: optimization objective, UAV dynamics constraint, and safety constraint.

Further, referring to FIG. 1 and FIG. 2 , the step S200 includes:

S210, defining an optimization target norm distance of the UAV system based on maximizing the performance of the original UAV flight control strategy while ensuring the safety of its output control action;

S220. Based on the received UAV state information and the predicted safety control action, the UAV dynamics transfer is performed to establish the UAV dynamics constraints of the safety control action within the limited time domain [0, M-1].

Further, referring to FIG. 1 and FIG. 2 , in step S210, the optimization target norm distance is:

in, is the unverified potentially unsafe control action output by the original UAV flight control strategy at time t, /> It is the safety control action of the UAV at time t obtained after safety verification and optimization of the correction method.

Specifically, the optimization goal is to optimize the safety verification and correction method proposed by the present invention, that is, to maximize the performance of the original UAV flight control strategy while ensuring the safety of its output control action, which is only related to the unverified potential unsafe control action. There is no need to use complex mathematical forms to construct it, just minimize the unsafe control action/> and predictive safety control actions/> The above optimization purpose can be achieved by simply calculating the norm distance between them.

Further, referring to FIGS. 1 and 2 , the unverified potentially unsafe control actions output by the original UAV flight control strategy for:

Among them, _Ft is the combined thrust generated by all the onboard motors in the drone, and _ωt is the three-axis angular velocity of the drone in the body coordinate system. represents the unsafe control action output by the original drone flight control strategy at time t, which is composed of the combined thrust _Ft generated by the four onboard motors of the drone and the three-axis angular velocity _ωt in the body coordinate system.

Further, referring to FIG. 1 and FIG. 2 , in step S220, for all integers m within a limited time domain, the UAV dynamics constraint is:

in, and m is an integer, /> represents all integer values 0, 1, ..., m, M-1 within a limited time domain; s _t+m is the state of the drone at time t+m, /> is the safety control action of the UAV at time t+m, and s _t+m+1 is the state at the next time t+m+1 calculated by the UAV dynamics model f(·). /> represents the safety control action of the drone at time t obtained after safety verification and optimization of the correction method. Similarly, /> Indicates the safety control action of the drone at time t+m.

UAV dynamic constraints: UAV dynamic constraints are an important basis for the optimization problem to transfer and update the UAV state within a limited time domain [0, M-1] according to the safety control actions obtained by real-time prediction. It is also a necessary condition to ensure that the UAV meets the safety constraints at all times in the future. Therefore, UAV dynamic transfer based on the received UAV state information and the predicted safety control actions is one of the constraints that need to be met in this optimization problem.

Further, referring to FIG. 1 and FIG. 2 , in step S220, the drone state s _t at time t includes:

s _t =[p _t ,q _t ,v _t ],

Among them, p _t is the position information of the UAV at time t, q _t is the attitude quaternion information of the UAV at time t, and v _t is the linear velocity information of the UAV at time t.

Similarly, the drone states at time t+m and t+m+1 are represented as s _t+m , s _t+m+1 respectively.

Further, referring to FIG. 1 and FIG. 2 , in step S300 , the safety constraint of the drone state includes that the drone state within a limited time domain must be within a safe state set S,

The security state set S is:

Among them, s _t is the state of the UAV at time t, B is the weight matrix of the state safety constraint on each state component, is the safety state boundary value that the drone state needs to meet at time t,

in, is the minimum value of the drone's position information at time t, /> is the minimum value of the UAV's attitude quaternion information at time t,/> is the minimum value of the UAV's linear velocity information at time t, /> is the maximum value of the drone's position information at time t, /> is the maximum value of the UAV's attitude quaternion information at time t,/> is the maximum value of the UAV’s linear velocity information at time t.

Further, referring to FIG. 1 and FIG. 2 , in step S300 , the safety constraints of the drone control actions include that the drone control actions within a limited time domain must be within the safety action set U,

The security action set U includes:

Among them, U represents the set of safety constraints that the drone control action needs to meet,

Where A represents the weight matrix of the control action safety constraint on each control component, is the safety boundary value that the drone control action needs to meet at time t,

in, is the minimum value of the combined thrust generated by all onboard motors in the drone,/> is the maximum value of the combined thrust generated by all the onboard motors in the drone,/> is the minimum value of the three-axis angular velocity of the drone in the body coordinate system, /> It is the maximum value of the three-axis angular velocity of the UAV in the body coordinate system.

Further, referring to FIG. 1 and FIG. 2 , at the end time M of the finite time domain, the drone state s _t+M needs to satisfy the state end set S _end , that is,

s _t+M ∈S _end ,

Among them, S _end is the state termination set, and the state termination set S _end is a subset of the state safety constraint set S, that is,

Among them, S is the set of state safety constraints.

Specifically, safety constraints are conditions that must be met to ensure that the drone state and control actions are always within a safe range. The drone state and control actions within a limited time domain must be within the safe state set S and safe action set U, that is, each state component p _t , q _t , v _t and control action component F _t , ω _t must satisfy the safety constraint boundary conditions, which can be specifically expressed as Among them, S _end is a subset of the set S, that is, Represents the state termination set of the UAV within a limited time domain. In order to ensure the recursive feasibility of solving the optimization problem, the state termination set S _end is defined as an invariant set, that is, there exists a control strategy/> For/> π(s)∈U and f(s,π(s))∈S _end hold, so that when the optimization problem has a feasible solution at the initial time t ₀ , it also has a feasible solution at the future time t＞t _0. In addition, in order to realize the adaptive ability of the safety verification and correction method to diversified scenarios, the present invention uses the state safety constraint boundary obtained from the environmental feature perception module in the drone system/> Real-time location information in/> The state safety set S in the above safety constraints is adjusted in real time so that it satisfies This makes it consistent with the position safety boundary conditions that the drone is currently in. Combining the components of the above optimization problem, by performing real-time and efficient solutions in a limited time domain, the drone control action that meets the safety constraints is finally obtained, achieving the purpose of safe flight control of the drone.

Further, referring to FIGS. 3 to 6 , the present invention also proposes a UAV control strategy safety verification and correction system for implementing a UAV control strategy safety verification and correction method, wherein the UAV control strategy safety verification and correction system comprises:

An unmanned aerial vehicle system, the unmanned aerial vehicle system including an existing unmanned aerial vehicle flight control strategy, the existing unmanned aerial vehicle flight control strategy generating unverified potentially unsafe control actions based on unmanned aerial vehicle state information of the unmanned aerial vehicle system to control the flight of the unmanned aerial vehicle system;

A safety verification and correction device, wherein the safety verification and correction device obtains potential unsafe control actions from the original UAV flight control strategy of the UAV system, generates safe control actions and transmits them to the UAV system, and the safety verification and correction device is connected to the UAV system.

Referring to FIG. 3 , in some specific embodiments, the simplest implementation scheme of the present invention mainly includes three parts, namely, the unmanned aerial vehicle system, the original unmanned aerial vehicle flight control strategy, and the safety verification and correction method. As the controlled object, the unmanned aerial vehicle system is mainly used to receive the safety control action generated by the control strategy after safety verification and correction, and the change and update of the unmanned aerial vehicle state information is generated by applying the control action to the bottom actuator of the system. At the same time, according to the changes in the environment around the unmanned aerial vehicle system, the safety constraint boundary is adaptively adjusted through the real-time environmental feature perception module, so that the safety verification and correction method has the adaptive ability of diversified scenarios. The original unmanned aerial vehicle flight control strategy generates unverified potential unsafe control actions based on the current unmanned aerial vehicle state information, which is used to realize the flight control of the unmanned aerial vehicle system. In order to ensure the safety of the control action acting on the unmanned aerial vehicle system, it is necessary to use the safety verification and correction method to detect the potential unsafe control action, and to correct it for safety based on the detection result. In the safety verification and correction method, the optimization problem is configured using the unsafe control action input at the same time, the current unmanned aerial vehicle system state information, and the adaptively adjusted safety constraint boundary, and the finite time domain optimization problem with the unmanned aerial vehicle system dynamics and safety constraints is solved to obtain the safety control action for the flight control of the unmanned aerial vehicle system. Based on the mutual feedback and combination of the three parts in the above-mentioned simplest implementation scheme, through the continuous transmission and interaction between the drone system status information and control actions, such a cycle is finally implemented, which ultimately ensures that the drone system is always in a safe flight state.

Referring to FIG4, in some specific embodiments, the UAV system, as the subject of the controlled object, is mainly used to receive the control action generated by the UAV flight control strategy, and inputs its own real-time state information as feedback to the UAV flight control strategy, forming a closed-loop UAV control system to achieve the purpose of controlling the UAV system. Referring to FIG4, the UAV system is mainly composed of two parts, namely, an onboard computer and an underlying flight controller. The onboard computer uses the sensors carried by the UAV system to obtain the characteristic information of the environment in which the UAV is located and the state information of the UAV system itself through the environmental feature perception module and the positioning module. After obtaining the characteristic information of the environment in which the UAV is located, the safety constraint boundary to be satisfied by the UAV flight control strategy can be adaptively adjusted based on the information, so that the corresponding safety verification and correction method has the adaptive ability of diversified scenarios, thereby ensuring that the UAV system can obtain the corresponding safety control action in real time in a complex and changeable environment, which is used to realize its safe flight control. After obtaining the state information of the UAV system itself, it is fused with the UAV attitude information measured by the inertial measurement unit in the underlying flight controller to obtain more accurate UAV system state estimation information, which is used as the input of the UAV flight control strategy and the corresponding safety verification and correction method, and becomes a necessary condition for obtaining safety control actions. After obtaining the safety control action, the underlying flight controller performs underlying angular velocity control on the UAV system based on the current UAV attitude information obtained by the inertial measurement unit, and finally obtains the control quantity used to drive the actuator, thereby achieving overall control of the UAV system.

Referring to FIG. 5 , in some specific embodiments, the original UAV flight control strategy refers to a UAV flight control strategy that has not been safety verified and has potential safety hazards. According to the unified and universal characteristics of the UAV flight control strategy safety verification and correction method proposed in the present invention, the original UAV flight control strategy does not specifically refer to a certain UAV flight control strategy, among which there are N possible control strategies. The present invention lists two more common ones, namely traditional control strategy and reinforcement learning control strategy. The original UAV flight control strategy is based on the UAV state information obtained in the UAV system. After solving different control strategy algorithms, the control action for the flight control of the UAV system can be obtained. However, due to the potential safety hazards of the original UAV flight control strategy, the output UAV control action also has certain safety risks. Therefore, before acting on the actual UAV system, it needs to be tested and adjusted by the safety verification and correction method to ensure that it outputs safe UAV control actions.

7 , in some specific embodiments, in order to solve the above optimization problem, at time t, the unsafe control action output by the original UAV flight control strategy is UAV state information s _t and state safety constraint boundary information/> At the same time, it is input into the optimization problem. By looping and iterating M times in the limited time domain range [0, M-1], the safety state sequence of the UAV in the next M time steps s _t+1 , s _t+2 ,…, s _t+M and the corresponding safety control action sequence can be predicted/> Take the safety control action corresponding to time t/> As an alternative to the original UAV flight control action/> The correction result can ensure that the UAV is always in a safe flight state, thereby achieving the purpose of safe flight control of the UAV.

For the specific implementation process, the present invention uses the unsafe reinforcement learning (RL) control strategy as the original flight control strategy of the UAV. On this basis, the proposed safety verification and correction method (Adaptive Safety Predictive Corrector, ASPC) with environmental adaptive capability is used to verify its safety. The UAV safe flight control simulation experiment is implemented in a hall environment with limited height, a corridor environment with limited width, and a room environment with limited space. The specific parameter configuration required for solving the optimization problem of the safety verification and correction method is shown in Table 1:

Table 1 Parameter configuration of optimization problem for security verification and correction method

Referring to FIG8 , for the simulation results, during the UAV simulation experiment, the unsafe control action (RL) output by the original UAV flight control strategy and the safe control action (Safe) data optimized by the safety verification and correction method (ASPC) can be displayed in real time.

From the data displayed in the simulation process, it can be seen that the safety verification and correction method proposed in the present invention has real-time and efficient solution capabilities, and the single solution time is within 2ms, which meets the real-time requirements of UAV flight control. In addition, by recording the flight trajectory data of the UAV and the original control actions and the corrected safety control action sequence at the corresponding time during flight, the obtained data is plotted into simulation results.

Referring to Figures 9 to 11b, Figure 9 shows the take-off test results of the drone in a hall environment with limited height, and the height and control action of the drone are always within the safe range; Figure 10 shows the crossing test results of the drone in a corridor environment with limited width, and the position of the drone is always within the safe boundary; Figures 11a and 11b respectively show the trajectory tracking test results of the drone in a room environment with limited space, and the flight trajectory and control action sequence of the drone during the trajectory tracking process show that it is always in a safe flight state. From the test results shown in the above figures, it can be seen that when the drone performs different flight missions in different space-constrained environments, it can always remain in the safe state space, thereby achieving the safe flight control effect of the drone, verifying the effectiveness of the safety verification and correction method proposed in the present invention and its adaptability to diverse environments. By comparing the difference between the drone control action before and after the safety verification in Figures 9 and 11b, it can be seen that whenever the drone approaches the safety constraint boundary, the safety verification and correction method proposed in the present invention will correct the unsafe control action output by the original drone flight control strategy in real time to ensure that the drone receives a safe control action, thereby achieving the purpose of safe flight control of the drone. In addition, when the UAV was not close to the safety constraint boundary, the safety verification and correction method did not make unnecessary corrections to the original UAV flight control strategy, thereby further verifying its optimization goal principle of maximizing the retention of the original UAV flight control strategy performance.

At present, most of the safe flight control strategies of UAVs are based on the idea of optimal control, and the purpose of safe flight control of UAVs is achieved by constructing complex optimization objective functions and safety constraints. However, in this process, it is necessary to rely on strict and complex mathematical forms to construct the optimization objectives and constraints of the UAV control strategy, which leads to its complex nonlinear structure and is often difficult to obtain its feasible solution through approximate solution. In addition, the satisfaction of the complex constraints of the UAV control strategy will continuously increase the conservatism of its output control actions, making it unsuitable for the flight maneuverability requirements of the UAV system. Although the model-free safety reinforcement learning method can train UAV flight control strategies that meet safety constraints through powerful strategy search and optimization capabilities, such methods will affect the efficiency of strategy exploration and optimization to a certain extent due to the limitations of safety constraints, and the trained UAV safety control strategies are deeply coupled with the application scenarios and specific flight tasks, and do not have universality and generalization capabilities for diversified application scenarios. In addition, during the training process of the UAV control strategy, due to the continuous trial and error and exploration of its interaction with the environment, it is impossible to avoid violating safety constraints, and thus its safety during the training process cannot be guaranteed. Safety verification-based methods are another optional method for improving the safety of UAV flight control. Although they can predict and correct unsafe control strategies through safety verification mechanisms, they also have problems such as high construction difficulty, high computational complexity, and poor environmental adaptability, making them impossible to be deployed and applied in real time in UAV systems in real scenarios.

Aiming at the safety risks existing in the design, training, testing and deployment of UAV flight control strategies, the present invention discloses a safety verification and correction method and system, the main technical key points of which are:

(1) For all potentially unsafe UAV flight control strategies, a unified and universal plug-and-play safety verification and correction method is proposed. The UAV flight control strategy is verified and corrected in a decoupled manner, thus achieving the purpose of safe UAV flight control.

(2) This safety verification and correction method is based on optimization theory. By designing specific optimization objectives and safety constraints, it obtains a safe flight control strategy through real-time and efficient calculation while maximizing the performance of the original UAV flight control strategy, thereby ensuring the safety of UAV flight control.

(3) This safety verification and correction method can adaptively adjust its safety constraint boundaries based on the current environmental feature information, so that it has the ability to adapt to diverse environments and is suitable for safe flight control of drones in multiple scenarios.

(4) In addition, based on the proposed UAV control strategy safety verification and correction method, the present invention designs a UAV system corresponding to the method to deploy and verify the feasibility and effectiveness of the proposed method.

Further, the present invention also proposes a computer-readable storage medium, on which program instructions are stored, and the described method is implemented when the program instructions are executed by a processor. It should be appreciated that the method steps in the embodiments of the present invention can be implemented or implemented by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-temporary computer-readable memory. The method can use standard programming techniques. Each program can be implemented in a high-level process or object-oriented programming language to communicate with a computer system. However, if necessary, the program can be implemented in assembly or machine language. In any case, the language can be a compiled or interpreted language. In addition, the program can be run on a programmed application-specific integrated circuit for this purpose.

Furthermore, the operations of the processes described herein may be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The processes described herein (or variations and/or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions, and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) that is executed collectively on one or more processors, by hardware, or a combination thereof. The computer program includes a plurality of instructions that may be executed by one or more processors.

Further, the method can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, etc. Various aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, an optical read and/or write storage medium, a RAM, a ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the process described herein. In addition, the machine-readable code, or portions thereof, can be transmitted via a wired or wireless network. When such media includes instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor, the invention described herein includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques of the present invention, the present invention can also include the computer itself.

The computer program can be applied to input data to perform the functions described herein, thereby converting the input data to generate output data stored in a non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the converted data represents physical and tangible objects, including specific visual depictions of physical and tangible objects produced on the display.

The above is only a preferred embodiment of the present invention. The present invention is not limited to the above implementation. As long as the technical effect of the present invention is achieved by the same means, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of protection of the present invention. Within the scope of protection of the present invention, its technical scheme and/or implementation method may have various modifications and changes.

Claims (10)

1. A method for verifying and correcting the safety of an unmanned aerial vehicle control strategy, wherein the method is applied to an unmanned aerial vehicle system, wherein the unmanned aerial vehicle system includes an existing unmanned aerial vehicle flight control strategy, and wherein the method comprises the following steps: S100, obtaining the current UAV system status information, the unverified potential unsafe actions output by the original UAV flight control strategy of the UAV system, and the safety constraint boundary of the UAV adaptive adjustment; S200, based on the optimization target norm, establish and solve a finite time domain optimization problem with UAV dynamics and safety constraints, and obtain a safety control action for UAV system flight control; S300, based on the safety constraint boundary of the UAV adaptive adjustment, establish and update the safety constraints that the UAV state needs to meet; S400, iteratively calculating step S200 and step S300 within a limited time domain to obtain a safety state sequence and a corresponding safety control action sequence of the UAV within a future limited time domain. 2. The method for verifying and correcting the safety of a drone control strategy according to claim 1, wherein step S200 comprises: S210, defining an optimization target norm distance of the UAV system based on maximizing the performance of the original UAV flight control strategy while ensuring the safety of its output control action; S220. Based on the received UAV state information and the predicted safety control action, the UAV dynamics transfer is performed to establish the UAV dynamics constraints of the safety control action within the limited time domain [0, M-1]. 3. The method for verifying and correcting the safety of the drone control strategy according to claim 2, characterized in that in step S210, the optimization target norm distance is: in, is the unverified potentially unsafe control action output by the original UAV flight control strategy at time t, /> It is the safety control action of the UAV at time t obtained after safety verification and optimization of the correction method. 4. The method for verifying and correcting the safety of the UAV control strategy according to claim 3 is characterized in that the unverified potentially unsafe control actions output by the original UAV flight control strategy for: Among them, _Ft is the combined thrust generated by all the onboard motors in the UAV, and _ωt is the three-axis angular velocity of the UAV in the body coordinate system. 5. The method for safety verification and correction of drone control strategy according to claim 2, characterized in that in step S220, for all integers m within a limited time domain, the drone dynamics constraint is: in, And m is an integer, s _t+m is the state of the drone at time t+m, /> is the safety control action of the UAV at time t+m, and s _t+m+1 is the state at the next time t+m+1 calculated by the UAV dynamics model f(·). 6. The method for verifying and correcting the safety of the drone control strategy according to claim 5, characterized in that, in the step S220, the drone state s t at time _t includes: s _t =[p _t ,q _t ,v _t ], Among them, p _t is the position information of the UAV at time t, q _t is the attitude quaternion information of the UAV at time t, and v _t is the linear velocity information of the UAV at time t. 7. The method for safety verification and correction of drone control strategy according to claim 1 is characterized in that in step S300, the safety constraint of the drone state includes that the state of the drone within a limited time domain must be within the safe state set S, The security state set S is: Among them, s _t is the state of the UAV at time t, B is the weight matrix of the state safety constraint on each state component, is the safety state boundary value that the drone state needs to meet at time t, in, is the minimum value of the drone's position information at time t, /> is the minimum value of the UAV's attitude quaternion information at time t,/> is the minimum value of the UAV's linear velocity information at time t, /> is the maximum value of the drone's position information at time t, /> is the maximum value of the UAV's attitude quaternion information at time t,/> is the maximum value of the UAV’s linear velocity information at time t. 8. The method for safety verification and correction of drone control strategy according to claim 1 is characterized in that in step S300, the safety constraints of the drone control action include that the drone control actions within a limited time domain must be within the safety action set U, The security action set U includes: Among them, U represents the set of safety constraints that the drone control action needs to meet, Where A represents the weight matrix of the control action safety constraint on each control component, is the safety boundary value that the drone control action needs to meet at time t, in, is the minimum value of the combined thrust generated by all onboard motors in the drone,/> is the maximum value of the combined thrust generated by all the onboard motors in the drone,/> is the minimum value of the three-axis angular velocity of the drone in the body coordinate system, It is the maximum value of the three-axis angular velocity of the UAV in the body coordinate system. 9. The method for verifying and correcting the safety of the UAV control strategy according to claim 8 is characterized in that: At the end time M of the finite time domain, the drone state s _t+M needs to satisfy the state termination set S _end , that is, s _t+M ∈S _end , Among them, S _end is the state termination set, and the state termination set S _end is a subset of the state safety constraint set S, that is, Among them, S is the set of state safety constraints. 10. A UAV control strategy safety verification and correction system, used to implement the method according to any one of claims 1 to 9, the UAV control strategy safety verification and correction system comprising: An unmanned aerial vehicle system, the unmanned aerial vehicle system including an existing unmanned aerial vehicle flight control strategy, the existing unmanned aerial vehicle flight control strategy generating unverified potentially unsafe control actions based on unmanned aerial vehicle state information of the unmanned aerial vehicle system to control the flight of the unmanned aerial vehicle system; A safety verification and correction device, wherein the safety verification and correction device obtains potential unsafe control actions from the original UAV flight control strategy of the UAV system, generates safe control actions and transmits them to the UAV system, and the safety verification and correction device is connected to the UAV system.