CN107943066B - Method for supervising and controlling obstacle avoidance of unmanned aerial vehicle by using human - Google Patents

Abstract

The present invention provides a method for supervising and controlling obstacle avoidance of unmanned aerial vehicles by manned aircraft, and relates to the field of unmanned aerial vehicles. The invention defines environmental factors, obstacle avoidance actions and values of supervisory control modes. Actions are reasoned and judged, and judgments are made according to the judgment results, so as to avoid obstacles. In the present invention, for known obstacles, the UAV fully exerts its autonomous ability and completes obstacle avoidance independently; for unknown obstacles, the UAV must be under the guidance of a human-machine. To complete the avoidance, the variable autonomous supervision and control method gives full play to the autonomous execution of the UAV and the analysis and judgment ability of the manned operator; for extreme situations such as communication interruption, the UAV adjusts the supervision and control mode and waits for the communication to resume. At the same time to ensure its own safety, the variable autonomous supervision and control mode combines the characteristics of unmanned aerial vehicles and manned aircraft, and has better adaptability to different types of obstacles and environmental conditions.

Description

A Supervision and Control Method for UAV Obstacle Avoidance by Manned Aircraft

technical field

The invention relates to the field of unmanned aerial vehicles, in particular to an evasion judgment method.

Background technique

When the UAV faces obstacles during its travel, a manned operator interacts with the UAV to ensure that the UAV completes obstacle avoidance. Supervisory control refers to the intermittent interaction of manned aircraft with the UAV, receiving feedback and giving instructions to control the process of the UAV in the mission environment.

However, in the practical application of obstacle avoidance in UAV systems, the traditional methods generally use the manual operation of the operator to avoid obstacles, or use path planning, rules or automata to avoid obstacles by calculating the cost function. Two problems will arise in this way: 1) the operator's "overload" phenomenon due to the complexity of the environment; 2) because the UAV system is in a too high autonomous authority, the operator loses the "human" ability to sense the surrounding environment. Out-of-the-Loop (OOTL) phenomenon. Two questions indicate that the supervision and control of manned aircraft to avoid UAV obstacles needs to be dynamically adjusted according to the actual situation.

Therefore, it is very necessary to study an unmanned aerial vehicle that can give full play to the high-level cognitive decision-making ability of the manned operator and the UAV's ability to perform tasks efficiently, while maintaining the situational awareness of the manned operator's environment and appropriate workload. Human-machine obstacle avoidance supervisory control method.

SUMMARY OF THE INVENTION

In order to overcome the deficiencies of the prior art, the present invention proposes an obstacle avoidance supervision control method, which can not only make the advantages of man and machine complement each other, but also maintain the operator's ability to perceive the environmental situation and appropriate workload, and avoid the occurrence of problems in supervision and control. Two problems, so as to effectively achieve the purpose of obstacle avoidance.

The obstacle avoidance supervision and control method proposed by the present invention is a method in which manned and unmanned aerial vehicles adjust different supervisory control modes according to environmental changes, and the unmanned aerial vehicle completes corresponding obstacle avoidance actions, so as to achieve the purpose of obstacle avoidance.

The technical scheme adopted by the present invention to solve its technical problem comprises the following steps:

Step1: Define environmental factors, obstacle avoidance actions and supervisory control modes

When the drone encounters an obstacle, there are 4 kinds of obstacle avoidance actions, including autonomous obstacle avoidance action A1, deceleration waiting for command action A2, hovering waiting for command action A3 and return action A4;

Among them, the autonomous obstacle avoidance action A1 is defined as a high-level action; the deceleration waiting command A2 and the hovering waiting command A3 are defined as medium-level actions; return A4 is a low-level action;

There are 3 modes of supervision and control, and the autonomous ability of the UAV from high to low is the exception management mode L3, the consent management mode L2 and the command control mode L1; the initial mode of the UAV is the exception management mode L3;

The definition of UAV supervision and control mode is shown in Table 1;

Table 1 Three supervisory control modes and their corresponding actions

The action execution mode refers to the way that the UAV completes the corresponding obstacle avoidance action under the supervision and control of the human machine, as shown in Table 2:

Table 2 Three supervisory control modes and their corresponding action execution modes

1) Environmental factors include three aspects: the distance of obstacles, whether the obstacle information is known, and whether the current communication status is good. The definitions are as follows:

① Definition of obstacle distance, including obstacle distance near V3 and obstacle distance far V5;

The position of the obstacle relative to the UAV can be expressed as

In formula (1), P(t) is the position of the obstacle i at time t, t is the time, P(0) is the initial position of the obstacle i, v _i is the velocity of the obstacle i, and v _u (τ) is the speed of the drone u;

为无人机的偏航角，无人机与障碍物距离的远近由到达障碍物位置的时间T由如下的公式定义：In formula (2)

is the yaw angle of the UAV, the distance between the UAV and the obstacle is defined by the time T to reach the obstacle position by the following formula:

②Whether the obstacle is known is divided into known obstacle V1 and unknown obstacle V4. Known obstacle means that the image information and position information of obstacles in the work area are known, while unknown obstacles are the image information of obstacles in the work area. and at least one of the location information is unknown;

③The current communication status is divided into normal communication and interruption V2. Communication interruption refers to the situation that the signal strength reaching the UAV receiver is less than the sensitivity of the manned receiver, resulting in the failure of normal communication between the manned aircraft and the UAV;

2) The value range of the values of environmental factors and obstacle avoidance actions is {0,1}, where 1 indicates that the state is valid, and 0 indicates that the state is invalid; the values of specific environmental factors and action states are shown in Table 3 and Table 4 below. :

Table 3 Values of Environmental Factors

Among them, * represents an irrelevant item, and an irrelevant item means that the value can be "0" or "1", and the value does not affect the corresponding state under *, but only affects the state under which it needs to take a specific value;

Table 4 Values of the action state of the UAV

Step2: Reasoning and judgment of the optional obstacle avoidance action of the drone

1) Write out the state vector by Step1

The state vector is represented by C(k), where k represents the number of times, and C(0) represents the initial state vector. The state vector includes the entire content of obstacle avoidance actions and environmental factors. The format is:

Among them, when there is no evasive action, the default values of the drones A1 to A4 are all 0, that is,

V1~V5 are the values obtained in Step1, and are always fixed values in Step2;

2) Use the state vector C(k) for inference, the inference process is as follows:

① Multiply the adjacency weight matrix W by the state vector C(k) to obtain the intermediate vector X(k):

X(k)=C(k)W (6)

Among them, the adjacency weight matrix W between nodes is as follows:

②Use the state transition function f(x) to process each component x of the intermediate vector X(k) for the intermediate vector X(k). The state transition function f(x) is a binary step function:

Among them, x is the component of the intermediate vector X(k), the dimension of the vector X(k) is 4, and the value range of the component x is {0,1};

③Update the state vector C(k) with the intermediate vector X(k), that is

④Continuously repeat steps ①, ② and ③, that is, continuously update the state vector with C(k) until the state vector C(k+n+1)=C(k+n), where n represents the number of times, that is, the first The value of the (k+n+1)th state vector is the same as the value of the (k+n)th state vector;

其中a1,a2,a3,a4分别表示障碍规避动作A1,A2,A3,A4对应的取值，值为0或者1，a1、a2、a3、a4中取值为1的项，对应障碍规避动作为可选障碍规避动作；

Among them, a1, a2, a3, and a4 represent the corresponding values of obstacle avoidance actions A1, A2, A3, and A4, respectively, and the value is 0 or 1. The value of a1, a2, a3, and a4 is 1, which corresponds to the obstacle avoidance action. for optional obstacle avoidance actions;

⑤The drone provides optional obstacle avoidance action results to the manned aircraft;

Step3: Final obstacle avoidance action judgment

1) When one of the following conditions a) and b) is satisfied, the drone needs to interact with the human-machine:

a) The environmental factor contains unknown obstacles, and the value of the V4 item is "1", that is, V=(*,*,*,1,*), where * represents an irrelevant item, and the irrelevant item represents a value of "0" or "1" ";

b) The optional obstacle avoidance action does not conform to the action level corresponding to the current supervision and control mode of the UAV;

The actions performed in the current supervisory control mode are as follows:

①When the supervisory control mode of the UAV is the exception management mode L3, and the UAV needs to interact, if the UAV cannot provide the obstacle avoidance action reasoning result within 15s, that is, the obstacle is unknown, the UAV will reduce the supervisory control mode to Agree to management mode L2; if the UAV provides the obstacle avoidance action reasoning result to the manned aircraft, the UAV adopts the intelligent combination method to actively feedback, and the operator executes the obstacle avoidance action reasoning result provided by the UAV within 15s without denying it, otherwise , the operator will not execute if it is negative within 15s, and reduce the supervisory control mode to the consent management mode L2;

② When the supervision and control mode of the UAV is the consent management mode L2 and the UAV needs to interact, if the UAV cannot provide the obstacle avoidance action reasoning result within 15s, that is, the obstacle is unknown, the UAV will reduce the supervision control mode to Command control mode L1; if the UAV provides the obstacle avoidance action reasoning result to the manned aircraft, the UAV adopts the intelligent combination method to actively feedback, waiting for the operator's approval, the operator will execute the obstacle provided by the UAV if the operator does not deny it within 15s Avoid the action inference result, if it is not negative within 15s, it will not be executed, and reduce the supervisory control mode to the instruction control mode L1;

③ When the supervisory control mode of the UAV is the command control mode L1, and the UAV needs to interact, the UAV provides the image information and position information of the obstacle to the manned aircraft, and provides the state vector C(k) and the obstacle avoidance action. As a result of the inference, the man-machine takes over to execute the action:

if (obstacle type is known), then (choose autonomous obstacle avoidance decision avoidance)

if (the type of obstacle is unknown and the distance is far), then (select the deceleration waiting command) (9)

if (the type of obstacle is unknown and the distance is close), then (choose to hover and wait for the command)

Give the UAV obstacle avoidance action by command. The UAV waits for the operator's decision within 15s, and executes according to the decision result. If there is no decision after timeout, it will return to the base;

2) When the optional obstacle avoidance action conforms to the action level corresponding to the current supervision and control mode of the UAV, and the environmental factors do not appear in the judgment condition a) of step 1) in step 2, the UAV does not need to interact with the human-machine, and can The selected obstacle avoidance actions are executed according to the following priorities:

A1>A2>A3>A4 (10)

The action execution methods in the corresponding supervisory control mode are as follows:

①When the drone does not need interaction, when in the exception management mode L3, the allowed actions are autonomous obstacle avoidance action A1, deceleration waiting command action A2, hovering waiting command action A3 and return action A4. The drone follows the formula (10) The priority adopts the autonomous execution method;

② When the drone does not need interaction, when in the consent management mode L2, the allowed actions are deceleration waiting for command action A2, hovering waiting for command action A3, and return action A4. The drone adopts the autonomous priority according to formula (10). execution way;

③ When the drone does not need interaction, when in the command control mode L1, the allowed action is the return action A4; the drone adopts the autonomous execution method according to the priority of formula (10);

Step4: Update the values of the environmental factors and obstacle avoidance actions of Step1 every 30s. If the UAV supervision control mode is lowered, update the UAV supervision control mode to the reduced mode. Repeat Step1 to Step3, and the UAV executes The final obstacle avoidance action is completed, and after the obstacle avoidance is successfully completed, the UAV supervision and control mode returns to the exception management mode L3.

The beneficial effect of the invention is that for known obstacles, the UAV can fully exert its autonomous ability and complete obstacle avoidance independently; for unknown obstacles, the UAV must complete the avoidance under the guidance of a manned aircraft, and the variable autonomous supervision and control method can Give full play to the autonomous execution of the UAV and the analysis and judgment ability of the manned operator; for extreme situations such as communication interruption, the UAV can ensure its own safety while waiting for the restoration of communication by adjusting the supervision and control mode. The variable autonomous supervision control mode can combine the characteristics of unmanned aerial vehicles and manned aircraft, and has better adaptability to different obstacle types and environmental conditions.

Description of drawings

Fig. 1 is a model block diagram of the UAV obstacle avoidance of the present invention.

FIG. 2 is a process diagram of obstacle avoidance in Embodiment 1 of the present invention.

FIG. 3 is a diagram of an obstacle avoidance process according to Embodiment 2 of the present invention.

FIG. 4 is a diagram of an obstacle avoidance process according to Embodiment 3 of the present invention.

Detailed ways

The present invention will be further described below with reference to the accompanying drawings and embodiments. FIG. 1 is a model block diagram of the UAV obstacle avoidance of the present invention.

Step1: Define environmental factors, obstacle avoidance actions and supervisory control modes

When the drone encounters an obstacle, there are 4 kinds of obstacle avoidance actions, including autonomous obstacle avoidance action A1, deceleration waiting for command action A2, hovering waiting for command action A3 and return action A4;

Among them, the autonomous obstacle avoidance action A1 is defined as a high-level action; the deceleration waiting command A2 and the hovering waiting command A3 are defined as medium-level actions; return A4 is a low-level action;

There are 3 modes of supervision and control, and the autonomous ability of the UAV from high to low is the exception management mode L3, the consent management mode L2 and the command control mode L1; the initial mode of the UAV is the exception management mode L3;

The definition of UAV supervision and control mode is shown in Table 1;

Table 1 Three supervisory control modes and their corresponding actions

The action execution mode refers to the way that the UAV completes the corresponding obstacle avoidance action under the supervision and control of the human machine, as shown in Table 2:

Table 2 Three supervisory control modes and their corresponding action execution modes

1) Environmental factors include three aspects: the distance of obstacles, whether the obstacle information is known, and whether the current communication status is good. The definitions are as follows:

① Definition of obstacle distance, including obstacle distance near V3 and obstacle distance far V5;

The position of the obstacle relative to the UAV can be expressed as

In formula (1), P(t) is the position of the obstacle i at time t, t is the time, P(0) is the initial position of the obstacle i, v _i is the velocity of the obstacle i, and v _u (τ) is the speed of the drone u;

为无人机的偏航角，无人机与障碍物距离的远近由到达障碍物位置的时间T由如下的公式定义：In formula (2)

is the yaw angle of the UAV, the distance between the UAV and the obstacle is defined by the time T to reach the obstacle position by the following formula:

②Whether the obstacle is known is divided into known obstacle V1 and unknown obstacle V4. Known obstacle means that the image information and position information of obstacles in the work area are known, while unknown obstacles are the image information of obstacles in the work area. and at least one of the location information is unknown;

③The current communication status is divided into normal communication and interruption V2. Communication interruption refers to the situation that the signal strength reaching the UAV receiver is less than the sensitivity of the manned receiver, resulting in the failure of normal communication between the manned aircraft and the UAV;

2) The value range of the values of environmental factors and obstacle avoidance actions is {0,1}, where 1 indicates that the state is valid, and 0 indicates that the state is invalid; the values of specific environmental factors and action states are shown in Table 3 and Table 4 below. :

Table 3 Values of Environmental Factors

Among them, * represents an irrelevant item, and an irrelevant item means that the value can be "0" or "1", and the value does not affect the corresponding state under *, but only affects the state under which it needs to take a specific value;

Table 4 Values of the action state of the UAV

Step2: Reasoning and judgment of the optional obstacle avoidance action of the drone

1) Write out the state vector by Step1

The state vector is represented by C(k), where k represents the number of times, and C(0) represents the initial state vector. The state vector includes the entire content of obstacle avoidance actions and environmental factors. The format is:

Among them, when there is no evasive action, the default values of the drones A1 to A4 are all 0, that is,

V1~V5 are the values obtained in Step1, and are always fixed values in Step2;

2) Use the state vector C(k) for inference, the inference process is as follows:

① Multiply the adjacency weight matrix W by the state vector C(k) to obtain the intermediate vector X(k):

X(k)=C(k)W (6)

Among them, the adjacency weight matrix W between nodes is as follows:

②Use the state transition function f(x) to process each component x of the intermediate vector X(k) for the intermediate vector X(k). The state transition function f(x) is a binary step function:

Among them, x is the component of the intermediate vector X(k), the dimension of the vector X(k) is 4, and the value range of the component x is {0,1};

③Update the state vector C(k) with the intermediate vector X(k), that is

④Continuously repeat steps ①, ② and ③, that is, continuously update the state vector with C(k) until the state vector C(k+n+1)=C(k+n), where n represents the number of times, that is, the first The value of the (k+n+1)th state vector is the same as the value of the (k+n)th state vector;

其中a1,a2,a3,a4分别表示障碍规避动作A1,A2,A3,A4对应的取值，值为0或者1，a1、a2、a3、a4中取值为1的项，对应障碍规避动作为可选障碍规避动作。

Among them, a1, a2, a3, and a4 represent the corresponding values of obstacle avoidance actions A1, A2, A3, and A4, respectively, and the value is 0 or 1. The value of a1, a2, a3, and a4 is 1, which corresponds to the obstacle avoidance action. For optional obstacle avoidance actions.

⑤The drone provides optional obstacle avoidance action results to the manned aircraft;

Step3: Final obstacle avoidance action judgment

1) When one of the following conditions a) and b) is satisfied, the drone needs to interact with the human-machine:

a) The environmental factor contains unknown obstacles, and the value of the V4 item is "1", that is, V=(*,*,*,1,*), where * represents an irrelevant item, and the irrelevant item represents a value of "0" or "1" ";

b) The optional obstacle avoidance action does not conform to the action level corresponding to the current supervision and control mode of the UAV;

The actions performed in the current supervisory control mode are as follows:

①When the supervisory control mode of the UAV is the exception management mode L3, and the UAV needs to interact, if the UAV cannot provide the obstacle avoidance action reasoning result within 15s, that is, the obstacle is unknown, the UAV will reduce the supervisory control mode to Agree to management mode L2; if the drone provides the obstacle avoidance action reasoning result to the manned drone, the drone adopts an intelligent combination method to actively feedback, and the operator executes the obstacle avoidance action reasoning result provided by the drone within 15s without denying it, otherwise , the operator will not execute if it is negative within 15s, and reduce the supervisory control mode to the consent management mode L2;

② When the supervision and control mode of the UAV is the consent management mode L2 and the UAV needs to interact, if the UAV cannot provide the obstacle avoidance action reasoning result within 15s, that is, the obstacle is unknown, the UAV will reduce the supervision control mode to Command control mode L1; if the UAV provides the obstacle avoidance action reasoning result to the manned aircraft, the UAV adopts the intelligent combination method to actively feedback, waiting for the operator's approval, the operator will execute the obstacle provided by the UAV if the operator does not deny it within 15s Avoid the action inference result, if it is negative within 15s, it will not be executed, and reduce the supervisory control mode to the operator decision mode L1;

③ When the supervisory control mode of the UAV is the operator decision mode L1, and the UAV needs to interact, the UAV provides the image information and position information of the obstacle to the manned aircraft, and provides the state vector C(k) and obstacle avoidance. As a result of the action inference, the man-machine takes over to execute the action:

Give the UAV obstacle avoidance action by command. The UAV waits for the operator's decision within 15s, and executes according to the decision result. If there is no decision after timeout, it will return to the base;

2) When the optional obstacle avoidance action conforms to the action level corresponding to the current supervision and control mode of the UAV, and the environmental factors do not appear in the judgment condition a) of step 1) in step 2, the UAV does not need to interact with the human-machine, and can The selected obstacle avoidance actions are executed according to the following priorities

A1>A2>A3>A4 (10)

The action execution methods in the corresponding supervisory control mode are as follows:

①When the drone does not need interaction, when in the exception management mode L3, the allowed actions are autonomous obstacle avoidance action A1, deceleration waiting command action A2, hovering waiting command action A3 and return action A4. The drone follows the formula (10) The priority adopts the autonomous execution method;

② When the drone does not need interaction, when in the consent management mode L2, the allowed actions are deceleration waiting for command action A2, hovering waiting for command action A3, and return action A4. The drone adopts the autonomous priority according to formula (10). execution way;

③ When the drone does not need interaction, when in the command control mode L1, the allowed action is the return action A4; the drone adopts the autonomous execution method according to the priority of formula (10);

Step4: Update the values of the environmental factors and obstacle avoidance actions of Step1 every 30s. If the UAV supervision control mode is lowered, update the UAV supervision control mode to the reduced mode. Repeat Step1 to Step3, and the UAV executes The final obstacle avoidance action is completed, and after the obstacle avoidance is successfully completed, the UAV supervision and control mode returns to the exception management mode L3.

The cases of UAVs in obstacle avoidance can be divided into two categories according to normal communication and communication interruption. At the same time, when the communication is normal, it can be divided into two situations: known obstacles and unknown obstacles. The following three cases are discussed separately. The process is simulated, as shown in Table 5:

Table 5 UAV obstacle avoidance

The embodiments shown in Table 5 are divided into two categories of normal communication and communication interruption, and normal communication is divided into two types: known obstacles and unknown obstacles. The simulation process of the task execution of the UAV under the three embodiments is described below:

Example 1

(1) Step1: Representation of environmental factors and obstacle avoidance actions

It can be seen from Table 5 that if the obstacle type is a known obstacle, then V1=1, V4=0; if the obstacle distance is a long distance, then V3=0, V5=1; if the communication status is normal, then V2=0, UAV Actions A1 to A4 are all initialized to 0, and the initial supervisory control mode is the exception management mode L3;

(2) Step2: Reasoning and judgment of optional obstacle avoidance actions

The state vector is:

Iterating it, the state vector of the output node is:

The output results show that the optional action of the UAV in Example 1 is autonomous obstacle avoidance A1, so for known obstacles, the UAV can fully exert its autonomous ability and complete obstacle avoidance independently.

(3) Step3: Final obstacle avoidance action judgment

For Example 1, the obstacle type is a known obstacle, the UAV has the capability of autonomous obstacle avoidance, and in the exception management mode L3, the UAV has the authority to perform autonomous obstacle avoidance actions, so there is no need to interact with the human-machine , the UAV can autonomously complete the obstacle avoidance process, as shown in Figure 2.

Case 2

(1) Step1: Representation of environmental factors and obstacle avoidance actions

It can be seen from Table 5 that if the obstacle type is unknown obstacle, then V1=0, V4=1; if the obstacle distance is short, then V3=1, V5=0; if the communication status is normal, then V2=0, the drone is moving A1 to A4 are all initialized to 0, and the initial supervisory control mode is the exception management mode.

(2) Step2: Reasoning and judgment of optional obstacle avoidance actions

The state vector is:

Iterating it, the state vector of the output node is:

The output result shows that in Example 2, the UAV cannot avoid obstacles autonomously, so it executes the hovering waiting command action A3.

(3) Step3: Final obstacle avoidance action judgment

The UAV cannot complete obstacle avoidance autonomously, and needs to interact with the man-machine and return the image information and position information of the obstacle at the same time. The man-machine operator updates the obstacle data according to the obstacle information, that is, the unknown obstacle information is updated to known obstacle information.

Update the representation of environmental factors and obstacle avoidance actions in Step1. If the obstacle type is a known obstacle, then V1=1, V4=0; if the obstacle distance is short, then V3=1, V5=0; if the communication status is normal, then V2=0, UAV actions A1-A4 are all initialized to 0, and the supervisory control mode is the exception management mode L3.

Update the reasoning judgment of the optional obstacle avoidance action of Step2

The state vector is:

Iterating over it, the output state vector is:

The output result indicates that the action chosen by the UAV under scenario 2 is autonomous obstacle avoidance A1.

Update Step3: Final obstacle avoidance action judgment

For Example 2, the obstacle type is unknown obstacle, and the UAV does not have the ability to avoid obstacles autonomously, so it needs to interact with the man-machine; the man-machine receives the obstacle image information and position information returned by the UAV, and updates the obstacle information , so that the type becomes a known obstacle; the UAV then makes a reasoning and judgment of the optional obstacle avoidance action according to the current environmental state. At this time, the obstacle type is a known obstacle, and the UAV has the ability to avoid obstacles autonomously. And in the exception management mode L3, the UAV has the authority to perform autonomous obstacle avoidance actions, so it does not need to interact with the human-machine, and the UAV can complete the obstacle avoidance process autonomously. The process is shown in Figure 3.

Example 3

(1) Step1: Representation of environmental factors and obstacle avoidance actions

According to Example 3, if the obstacle type is unknown obstacle, then V1=0, V4=1; if the obstacle distance is long, then V3=0, V5=1; if the communication status is communication interruption, then V2=1, the drone moves A1 to A4 are all initialized to 0, and the supervisory control mode is the exception management mode L3.

(2) Step2: Reasoning and judgment of optional obstacle avoidance actions

The state vector is:

Iterating it, the state vector of the output node is:

The output result shows that in the case of Example 3, the UAV does not have the ability to avoid obstacles autonomously, so it slows down and waits for the command A2, and because the communication is interrupted, it cannot receive the command from the manned aircraft. In order to ensure its own safety, the execution returns. Operate A4.

(3) Step3: Final obstacle avoidance action judgment

The UAV cannot complete the obstacle avoidance decision autonomously, and at the same time, due to the interruption of communication, it cannot interact with the human-machine. At this time, the UAV chooses to adjust the supervisory control mode. After 15s, the supervisory control mode is reduced to the operator decision-making mode L2, and the communication is still continued for 15s thereafter. Interrupted, unable to interact, the drone reduces the supervision control mode to the command control mode L1, after 15s, the communication is still interrupted and cannot be interacted, the drone selects the self-protection behavior according to the command control mode L1, and returns to the base, that is, the return to the base action A4 , the process is shown in Figure 4.

Claims (1)

1. a kind of supervising and controlling method of manned aircraft to UAV obstacle avoidance, is characterized in that comprising the following steps: Step1: Define environmental factors, obstacle avoidance actions and supervisory control modes When the drone encounters an obstacle, there are 4 kinds of obstacle avoidance actions, including autonomous obstacle avoidance action A1, deceleration waiting for command action A2, hovering waiting for command action A3 and return action A4; Among them, the autonomous obstacle avoidance action A1 is defined as a high-level action; the deceleration waiting command A2 and the hovering waiting command A3 are defined as medium-level actions; return A4 is a low-level action; There are 3 modes of supervision and control, and the autonomous ability of the UAV from high to low is the exception management mode L3, the consent management mode L2 and the command control mode L1; the initial mode of the UAV is the exception management mode L3; The definition of UAV supervision and control mode is shown in Table 1; Table 1 Three supervisory control modes and their corresponding actions The action execution mode refers to the way that the UAV completes the corresponding obstacle avoidance action under the supervision and control of the human machine, as shown in Table 2: Table 2 Three supervisory control modes and their corresponding action execution modes

1) Environmental factors include three aspects: the distance of obstacles, whether the obstacle information is known, and whether the current communication status is good. The definitions are as follows: ① Definition of obstacle distance, including obstacle distance near V3 and obstacle distance far V5; The position of the obstacle relative to the UAV can be expressed as

In formula (1), P(t) is the position of the obstacle i at time t, t is the time, P(0) is the initial position of the obstacle i, v _i is the velocity of the obstacle i, and v _u (τ) is the speed of the drone u;

In formula (2)

is the yaw angle of the UAV, the distance between the UAV and the obstacle is defined by the time T to reach the obstacle position by the following formula:

②Whether the obstacle is known is divided into known obstacle V1 and unknown obstacle V4. Known obstacle means that the image information and position information of obstacles in the work area are known, while unknown obstacles are the image information of obstacles in the work area. and at least one of the location information is unknown; ③The current communication status is divided into normal communication and interruption V2. Communication interruption refers to the situation that the signal strength reaching the UAV receiver is less than the sensitivity of the manned receiver, resulting in the failure of normal communication between the manned aircraft and the UAV; 2) The value range of the values of environmental factors and obstacle avoidance actions is {0,1}, where 1 indicates that the state is valid, and 0 indicates that the state is invalid; the values of specific environmental factors and action states are shown in Table 3 and Table 4 below. : Table 3 Values of Environmental Factors Among them, * represents an irrelevant item, and an irrelevant item means that the value can be "0" or "1", and the value does not affect the corresponding state under *, but only affects the state under which it needs to take a specific value; Table 4 Values of the action state of the UAV

Step2: Reasoning and judgment of the optional obstacle avoidance action of the drone 1) Write out the state vector by Step1 The state vector is represented by C(k), where k represents the number of times, and C(0) represents the initial state vector. The state vector includes the entire content of obstacle avoidance actions and environmental factors. The format is:

Among them, when there is no evasive action, the default values of the drones A1 to A4 are all 0, that is,

V1~V5 are the values obtained in Step1, and are always fixed values in Step2; 2) Use the state vector C(k) for inference, the inference process is as follows: ① Multiply the adjacency weight matrix W by the state vector C(k) to obtain the intermediate vector X(k): X(k)=C(k)W (6) Among them, the adjacency weight matrix W between nodes is as follows:

②Use the state transition function f(x) to process each component x of the intermediate vector X(k) for the intermediate vector X(k). The state transition function f(x) is a binary step function:

Among them, x is the component of the intermediate vector X(k), the dimension of the vector X(k) is 4, and the value range of the component x is {0,1}; ③Update the state vector C(k) with the intermediate vector X(k), that is

④Continuously repeat steps ①, ② and ③, that is, continuously update the state vector with C(k) until the state vector C(k+n+1)=C(k+n), where n represents the number of times, that is, the first The value of the (k+n+1)th state vector is the same as the value of the (k+n)th state vector;

Among them, a1, a2, a3, and a4 represent the corresponding values of obstacle avoidance actions A1, A2, A3, and A4, respectively, and the value is 0 or 1. The value of a1, a2, a3, and a4 is 1, which corresponds to the obstacle avoidance action. for optional obstacle avoidance actions; ⑤The drone provides optional obstacle avoidance action results to the manned aircraft; Step3: Final obstacle avoidance action judgment 1) When one of the following conditions a) and b) is satisfied, the drone needs to interact with the human-machine: a) The environmental factor contains unknown obstacles, and the value of the V4 item is "1", that is, V=(*,*,*,1,*), where * represents an irrelevant item, and the irrelevant item represents a value of "0" or "1" "; b) The optional obstacle avoidance action does not conform to the action level corresponding to the current supervision and control mode of the UAV; The actions performed in the current supervisory control mode are as follows: ①When the supervisory control mode of the UAV is the exception management mode L3, and the UAV needs to interact, if the UAV cannot provide the obstacle avoidance action reasoning result within 15s, that is, the obstacle is unknown, the UAV will reduce the supervisory control mode to Agree to management mode L2; if the UAV provides the obstacle avoidance action reasoning result to the manned aircraft, the UAV adopts the intelligent combination method to actively feedback, and the operator executes the obstacle avoidance action reasoning result provided by the UAV within 15s without denying it, otherwise , the operator will not execute if it is negative within 15s, and reduce the supervisory control mode to the consent management mode L2; ② When the supervision and control mode of the UAV is the consent management mode L2 and the UAV needs to interact, if the UAV cannot provide the obstacle avoidance action reasoning result within 15s, that is, the obstacle is unknown, the UAV will reduce the supervision control mode to Command control mode L1; if the UAV provides the obstacle avoidance action reasoning result to the manned aircraft, the UAV adopts the intelligent combination method to actively feedback, waiting for the operator's approval, the operator will execute the obstacle provided by the UAV if the operator does not deny it within 15s Avoid the action inference result, if it is negative within 15s, it will not be executed, and reduce the supervisory control mode to the instruction control mode L1; ③ When the supervisory control mode of the UAV is the command control mode L1, and the UAV needs to interact, the UAV provides the image information and position information of the obstacle to the manned aircraft, and provides the state vector C(k) and the obstacle avoidance action. As a result of the inference, the man-machine takes over to execute the action: if (obstacle type is known), then (choose autonomous obstacle avoidance decision avoidance) if (the type of obstacle is unknown and the distance is far), then (select the deceleration waiting command) (9) if (the type of obstacle is unknown and the distance is close), then (choose to hover and wait for the command) Give the UAV obstacle avoidance action by command. The UAV waits for the operator's decision within 15s, and executes according to the decision result. If there is no decision after timeout, it will return to the base; 2) When the optional obstacle avoidance action conforms to the action level corresponding to the current supervision and control mode of the UAV, and the environmental factors do not appear in the judgment condition a) of step 1) in step 2, the UAV does not need to interact with the human-machine, and can The selected obstacle avoidance actions are executed according to the following priorities: A1>A2>A3>A4 (10) The action execution methods in the corresponding supervisory control mode are as follows: ①When the drone does not need interaction, when in the exception management mode L3, the allowed actions are autonomous obstacle avoidance action A1, deceleration waiting command action A2, hovering waiting command action A3 and return action A4. The drone follows the formula (10) The priority adopts the autonomous execution method; ② When the drone does not need interaction, when in the consent management mode L2, the allowed actions are deceleration waiting for command action A2, hovering waiting for command action A3, and return action A4. The drone adopts the autonomous priority according to formula (10). execution way; ③ When the drone does not need interaction, when in the command control mode L1, the allowed action is the return action A4; the drone adopts the autonomous execution method according to the priority of formula (10); Step4: Update the values of the environmental factors and obstacle avoidance actions of Step1 every 30s. If the UAV supervision control mode is lowered, update the UAV supervision control mode to the reduced mode. Repeat Step1 to Step3, and the UAV executes The final obstacle avoidance action is completed, and after the obstacle avoidance is successfully completed, the UAV supervision and control mode returns to the exception management mode L3.

Images