Binocular non-directional motion control method, device and storage medium for binocular robot

Abstract

The present invention relates to a binocular robot's binocular counter-directional motion control method, device, and storage medium. The method comprises: controlling two image sensors to independently counter-directionally move, continuously acquiring images captured by two image acquisition units, determining whether the captured images contain a target object, and if so, entering a target tracking mode; after entering the target tracking mode, if only one image captured by the image sensor contains the target object, controlling the rotation until both images captured by the image sensors contain the target object; then, based on the deviation between the position of the target object and the center of the camera's field of view, and the incremental deviation between the previous and next frames, obtaining control values ​​for the two image sensors in the horizontal and vertical directions; and distributing the obtained control values ​​for the two image sensors in the horizontal and vertical directions to each servo to obtain a control value for each servo. Compared with the prior art, the present invention has the advantages of high control sensitivity and low energy consumption.

Description

Technical Field

[0001] The present invention relates to robot control technology, and in particular to a binocular dysdirectional motion control method, device and storage medium for a binocular robot. Background Art

[0002] The existing control scheme for binocular, non-directional motion bionic eyes is relatively simple and the control algorithm is relatively simple. It can only achieve the basic movement of the binocular bionic eyes by controlling multiple servos, and realize periodic rotation according to pre-given instructions to monitor and search the surrounding environment. However, this control method is relatively simple and cannot autonomously perceive the surrounding environment, cannot meet the bionic eye's demand for intelligence, and has high energy consumption and low search efficiency.

[0003] In the existing technology, the bionic eye robot has low efficiency and high energy consumption in autonomous perception of the surrounding environment, which makes it difficult to meet the needs of more intelligent human-computer interaction and has great limitations. It is necessary to propose a smarter and more efficient bionic eye robot design solution. Summary of the Invention

[0004] The purpose of the present invention is to provide a binocular anisotropic motion control method, device and storage medium for a binocular robot, which makes the control more flexible and reduces energy consumption.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] A binocular out-of-direction motion control method for a binocular robot, comprising:

[0007] When entering the search mode, the two image sensors are controlled to move in opposite directions, and the images captured by the two image acquisition units are continuously acquired to determine whether the captured images contain the target object. If so, the target tracking mode is entered;

[0008] When entering the target tracking mode, if only one image sensor captures an image containing the target object, the image sensor that has not detected the target object is controlled to rotate in the same direction as the image sensor that has detected the target object until the images captured by both image sensors contain the target object;

[0009] When entering the target tracking mode, if the images captured by the two image sensors both contain the target object, the deviation between the target object's position and the center of the camera's field of view, as well as the deviation increment between the previous and next frames, are sequentially fuzzified, fuzzy inference is performed according to the fuzzy rule table, and defuzzification is performed to obtain the control amount of the two image sensors in the horizontal and vertical directions. The obtained control amount of the two image sensors in the horizontal and vertical directions is then distributed to each servo to obtain the control amount of each servo.

[0010] The process of generating the control amounts of the two image sensors in the horizontal and vertical directions includes:

[0011] Perform fuzzy processing on the input deviation and deviation increment, select 7 split points, denoted as NB, NM, NS, ZO, PS, PM, PB, and evenly split the deviation and deviation increment domain into eight intervals. At the same time, these 7 split points are used as fuzzy subsets.

[0012] The linear membership function is used to calculate the membership of the lateral quantities of the deviation and the deviation increment;

[0013] According to the fuzzy rule table, fuzzy reasoning is performed to calculate the lateral component U of the PID output parameter w and the longitudinal component U h The membership of its fuzzy subset is defuzzified, and the increment of the lateral component ΔU is obtained according to the centroid method. w and the increment of the longitudinal component ΔU h ;

[0014] Introducing coefficient λ to ΔU w and ΔU h Perform scaling and update the horizontal and vertical components of the PID output parameters;

[0015] The updated PID output parameters are input into the PID controller to obtain the control quantities for the two image sensors in the horizontal and vertical directions.

[0016] The membership values ​​of the lateral quantity of the deviation and the deviation increment are both two, and the membership value of the lateral quantity of the deviation is C i 、C i+1 , for C i 、C i+1 The membership degrees are:

[0017] (E w -C i ) / (C i+1 -C i )、(C i+1 -E w ) / (C i+1 -C i )

[0018] Where: E w is the lateral amount of deviation.

[0019] ΔU w The mathematical expression is:

[0020]

[0021] Among them: F i For Uw Fuzzy subset, M i For U w F i The degree of membership.

[0022] The updating process of the lateral component of the PID output parameter is:

[0023] U w (n) = U w (n-1)+λΔU w

[0024] Among them: U w (n) is the updated lateral component, U w (n-1) is the horizontal component before updating.

[0025] The process of generating the control quantity of each servo includes:

[0026] For the space directly in front of the fuselage, a two-dimensional rectangular coordinate system is established on a horizontal plane passing through the center of the camera and a vertical plane passing through the fuselage's central axis and perpendicular to the fuselage. The coordinate system is fixed to the fuselage, dividing the field of view into several areas. Among them, two image sensors are installed on the fuselage;

[0027] Determine the area and position of the target object within the horizontal field of view;

[0028] Assign control weights to the neck and eyes based on the target object's location. The eyes are the servos that control the two image sensors, and the neck is the servo that controls the fuselage.

[0029] If the target object is in the middle area of ​​the horizontal field of view, the neck does not need to be moved and the neck weight is 0. If the target object is on both sides of the horizontal field of view, the neck needs to be moved. The neck weight is determined according to the two deviation angles of the target in the camera field of view. At the same time, the control weight of the eyes is allocated according to the neck weight.

[0030] According to the assigned control weights, based on the obtained control amounts for the left and right eyes in the horizontal and vertical directions, they are distributed to each servo to obtain the control amount of each servo.

[0031] There are six steering gears in total, and the control quantity of each steering gear is as follows:

[0032]

[0033] Where: D out is the vector of all servo control quantities, D out1 is the control quantity of servo 1, D out2 is the control quantity of the second servo, D out3 is the control quantity of the servo 3, D out4 is the control quantity of the servo 4, Dout5 is the control quantity of the servo 5, D out6 is the control quantity of servo six, C out_leftw is the horizontal control amount of the left image sensor, C out_left h is the vertical control amount of the left image sensor, W neck w Control weight for the horizontal direction of the neck, C out_righth is the vertical control amount of the right image sensor, C out_right w is the horizontal control amount of the right image sensor, W neck h is the control weight of the neck in the vertical direction, W left h is the vertical control weight of the left image sensor, W left w is the horizontal control weight of the left image sensor, W right h is the vertical control weight of the right image sensor, W right w is the horizontal control weight of the right image sensor.

[0034] When only one image sensor captures an image containing the target object, the control weight of the neck is 0.

[0035] A target search device based on a binocular unidirectional motion robot comprises a memory, a processor, and a program stored in the memory. When the processor executes the program, the method described above is implemented.

[0036] A storage medium stores a program thereon, and when the program is executed, the method described above is implemented.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. It can enable the heterodox binocular robot to autonomously perceive the surrounding environment, search for targets of interest and align them.

[0039] 2. The head-eye distributor can delegate the tasks of searching for and aligning with the target to the eye mechanism as much as possible, making the control more flexible and reducing energy consumption. When the target is located on the side of the robot, the two end effectors of the robot will need to rotate in the same direction. At this time, adopting a strategy of prioritizing the rotation of the neck will effectively improve the efficiency of the target alignment.

[0040] 3. The fuzzy PID control algorithm used has a faster response, smaller overshoot, and shorter adjustment time than traditional PID control, which improves the superiority of the controller. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is the fuzzy PID control block diagram;

[0042] Figure 2 It is the fuzzy controller structure;

[0043] Figure 3 This is a schematic diagram of the area division of the horizontal plane of the space around the big-eye robot;

[0044] Figure 4 It is a mode switch-state converter;

[0045] Figure 5 This is the system architecture diagram;

[0046] Figure 6 This is a schematic diagram of hardware device connections;

[0047] Figure 7 This is the actual effect picture;

[0048] Figure 8 This is a detailed logic diagram of the system of this application. DETAILED DESCRIPTION

[0049] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0050] The present invention realizes autonomous perception and motion control of a binocular robot with different directions of motion, designs a control scheme for an active perception bionic eye robot, and designs a head-eye distributor according to the corresponding robot behavior logic, thus designing a complete robot control scheme, which has important reference value for the intelligent development of binocular bionic robots with different directions of motion.

[0051] The embedded system uses the NVIDIA JETSON AGX XAVIER chip with an ARM6 architecture. Its supporting JetPack components include TensorRT and cuDNN for deep learning applications, CUDA for GPU acceleration, a multimedia API package for camera applications and sensor development, and VisionWorks and OpenCV for machine vision. Both cameras are RER-USB3MP01H-LS29-pin USB cameras with an actual resolution of 640×480. The motors use the Fiter SMS series and SCS series servos. The power board uses the LM2596S digital display multi-channel switching power supply, which is connected to the power board's 6V and 12V power supplies to ensure the normal operation of the servos of each rated power. The hardware connection diagram is shown in the figure. Figure 6 .

[0052] Specifically, a binocular out-of-direction motion control method for a binocular robot includes:

[0053] When entering the search mode, the two image sensors are controlled to move in opposite directions, and the images captured by the two image acquisition units are continuously acquired to determine whether the captured images contain the target object. If so, the target tracking mode is entered;

[0054] When entering the target tracking mode, if only one image sensor captures an image containing the target object, the image sensor that has not detected the target object is controlled to rotate in the same direction as the image sensor that has detected the target object until the images captured by both image sensors contain the target object;

[0055] When entering the target tracking mode, if the images captured by the two image sensors both contain the target object, then the deviation between the position of the target object and the center of the camera field of view, as well as the deviation increment between the previous and next two frames, are calculated as follows: Figure 1 As shown in the figure, fuzzy processing is performed in sequence, fuzzy reasoning is performed according to the fuzzy rule table, and defuzzification is performed to obtain the control quantities of the two image sensors in the horizontal and vertical directions. According to the obtained control quantities of the two image sensors in the horizontal and vertical directions, they are distributed to each servo to obtain the control quantity of each servo. The fuzzy controller structure is shown in the figure. Figure 2 As shown;

[0056] It enables the heterodox binocular robot to autonomously perceive the surrounding environment, search for targets of interest and align them.

[0057] Specifically, the operating mode switches based on target detection results. Two image sensors, namely cameras, concurrently capture images and detect targets within them. The target detection information transmitted to the main controller includes the detected location and detection success. If neither eye detects the target, the system enters search mode, rapidly scanning the surrounding environment in opposite directions while the target detector remains operational.

[0058] When either eye detects the target, it switches from search mode to tracking mode. If both eyes detect the target, the control amount for each servo is calculated based on the deviation to keep both eyes aligned with the target. If only the left eye detects the target, the left eye remains aligned normally, while the right eye is guided to adjust its position in the direction of the target. The specific method is to give the right eye a small, pre-set control amount to bring it closer to the target. Adjusting the control amount can control the speed of this process until the target appears within the right eye's range. In addition, the driving effect of the neck joint on the posture of both eyes is also important to consider. Therefore, even if the right eye is fixed, it will be driven by the neck joint to move closer to the target.

[0059] The process of generating the control quantities of the two image sensors in the horizontal and vertical directions includes:

[0060] Perform fuzzy processing on the input deviation and deviation increment, select 7 split points, denoted as NB, NM, NS, ZO, PS, PM, PB, and evenly split the deviation and deviation increment domain into eight intervals. At the same time, these 7 split points are used as fuzzy subsets.

[0061] The linear membership function is used to calculate the membership of the horizontal quantity of the deviation and the deviation increment. The membership values ​​of the horizontal quantity of the deviation and the deviation increment are two each, and the membership value of the horizontal quantity of the deviation is C i 、C i+1 , for C i 、C i+1 The membership degrees are:

[0062] (E w -C i ) / (C i+1 -C i )、(C i+1 -E w ) / (C i+1 -C i )

[0063] Where: E w is the lateral amount of deviation.

[0064] According to the fuzzy rule table, fuzzy reasoning is performed to calculate the lateral component U of the PID output parameter w and the longitudinal component U h For the membership of its fuzzy subset, the two components including horizontal and vertical have their own fuzzy subsets. The method and steps are the same, just as above. And U and ΔU represent all pid parameters in a certain direction (K p ,Ki,K d and ΔK p , ΔK i , ΔK d ), defuzzify the membership result, and get ΔU according to the centroid method w and ΔU h , where ΔU w The mathematical expression is:

[0065]

[0066] Among them: F i For U w Fuzzy subset, M i For U w F i The degree of membership.

[0067] Introducing coefficient λ to ΔU w and ΔU hScaling is performed to update the horizontal and vertical components of the PID output parameters. The update process of the horizontal component of the PID output parameter is as follows:

[0068] U w (n) = U w (n-1)+λΔU w

[0069] Among them: U w (n) is the updated lateral component, U w (n-1) is the horizontal component before updating.

[0070] This method is also applicable to the update of the longitudinal component.

[0071] In addition, the generation process of the control quantity of each servo includes:

[0072] The updated PID output parameters are input into the PID controller to obtain the control quantities for the two image sensors in the horizontal and vertical directions.

[0073] For the space directly in front of the fuselage, a two-dimensional rectangular coordinate system is established on a horizontal plane passing through the center of the camera and a vertical plane passing through the fuselage's central axis and perpendicular to the fuselage. The coordinate system is fixed to the fuselage, dividing the field of view into several areas. Among them, two image sensors are installed on the fuselage;

[0074] Determine the area and position of the target object within the horizontal field of view. For details on the division of the area, see Figure 3 ;

[0075] Assign control weights to the neck and eyes based on the target object's location. The eyes are the servos that control the two image sensors, and the neck is the servo that controls the fuselage.

[0076] According to the assigned control weights, based on the obtained control amounts for the left and right eyes in the horizontal and vertical directions, they are distributed to each servo to obtain the control amount of each servo.

[0077] There are six servos in total, and the control quantity of each servo is as follows:

[0078]

[0079] Where: D out is the vector of all servo control quantities, D out1 is the control quantity of servo 1, D out2 is the control quantity of the second servo, D out3 is the control quantity of the servo 3, D out4 is the control quantity of the servo 4, D out5 is the control quantity of the servo 5, D out6 is the control quantity of servo six, Cout_left w is the horizontal control amount of the left image sensor, C out_left h is the vertical control amount of the left image sensor, W neck w Control weight for the horizontal direction of the neck, C out_right h is the vertical control amount of the right image sensor, C out_right w is the horizontal control amount of the right image sensor, W neck h is the control weight of the vertical direction of the neck, W left h is the vertical control weight of the left image sensor, W left w is the horizontal control weight of the left image sensor, W right h is the vertical control weight of the right image sensor, W right w is the horizontal control weight of the right image sensor.

[0080] As mentioned above, the head-eye distributor can delegate the tasks of searching for and aligning with the target to the eye mechanism as much as possible, making the control more flexible and reducing energy consumption. When the target is located on the side of the robot, the two end effectors of the robot will need to rotate in the same direction. At this time, the strategy of prioritizing the rotation of the neck will effectively improve the efficiency of aligning with the target.

[0081] See also Figure 5 , using a specific case to introduce the steps of implementing the control method:

[0082] a: In the robot system, it is initialized to the search mode. The binocular robot image sensors 1 and 2 move in opposite directions to collect images of the surrounding environment in parallel, and transmit the collected images to the target detection controller 3.

[0083] b: Based on the images transmitted by image sensors 1 and 2, the YOLOv5 algorithm is used for parallel target detection and the detection results are transmitted to the main controller in real time; sensor 2 successfully detects the target object and transmits its location to the main controller.

[0084] c: The state transfer device 4 in the main controller switches from the search mode to the alignment tracking mode according to the target detection result, and transmits the target position information to the fuzzy PID controller 5. Figure 4 and Figure 8 shown.

[0085] d: Fuzzy PID controller 5 calculates binocular vision deviation E and deviation increment E C , E w The range is [-320, 320], E h The range is [-240, 240], The range is [-64, 64], The range is [-48, 48], E w 、E h 、 They are the deviation E and the deviation increment E C Next, E and E C Perform fuzzy processing. w For example, its range is divided into 8 sub-intervals, namely [-320,-240], [-240,-160],

[0086] [-160,-80], [-80,0], [0,80], [80,160], [160,240], [240,320]. Take -240,-160,-80,0,80,160,240 as fuzzy subsets and denote them as C i , where i=1,2,3,…,7,

[0087] The linear membership function is used to calculate E w The membership degree of C i <E w <C i+1 , i=1,2,3,…6, then E w There are two membership values ​​C i 、C i+1 , for C i 、C i+1 The membership degrees are:

[0088] (E w -C i ) / (C i+1 -C i )、(C i+1 -E w ) / (C i+1 -C i )

[0089] The same calculation Degree of membership.

[0090] Perform fuzzy reasoning according to the corresponding fuzzy rule table and calculate the parameters of PID output using U w (Generally refers to the PID parameter lateral component, including ) for its fuzzy subset F i (For U w The membership of the domain is divided into eight intervals (7 values). The membership results are defuzzified and ΔU is obtained according to the centroid method. w (Transverse component U w The increment, including ΔK pw , ΔK iw , ΔK dw )

[0091]

[0092] Introducing coefficient λ to ΔU w Scaling (different ΔU w Can correspond to different λ), we can get

[0093] U w (n) = U w (n-1)+λΔU w

[0094] The longitudinal component U of the PID parameter can also be obtained h (n), so far the solution is obtained to obtain the 6 PID horizontal and vertical dimension parameters that change in real time according to the actual situation Then input it into the PID controller and the output result is a two-dimensional vector Used to adjust the posture of the left eye, and also get the corresponding output two-dimensional vector of the right eye The obtained control quantity is recorded as matrix C out , compared with the PID controller, the MAE value is reduced by 19.5%.

[0095]

[0096] e: The head-eye distributor 6 calculates the deviation angles α and β according to the pixel position of the target in the camera field of view, and determines the area within the field of view where the target is located. If the target is in the area on both sides, the head-neck weight is

[0097] W neckw is amax{-α,-β}

[0098] and Both

[0099] Where a is the proportionality coefficient, Control weights for the neck. They are the control weights of the left and right eye horizontal servos respectively.

[0100] Likewise, in the vertical direction, the neck controls the weight 0, left and right eyes control weights Each is 0.5. Considering that the neck servo will drive the two cameras to rotate at the same time, due to the short control cycle, the rotation amplitude of each servo is not large, and the neck effect is similar to the two eye servos rotating in the same direction and at the same angle, so the corresponding control amount is regarded as

[0101] The control quantity assigned to the servo is represented by a 6-dimensional vector D out The final control quantity of each servo is

[0102]

[0103] Among them D out1 , D out2 , D out3 , D out4 , D out5 , D out6 Respectively represent the control amount of servo 1 to servo 6, where D out1 、D out2 , represents the neck servo control amount, D out3 , D out4 , D out5 , D out6 It is the eye servo control quantity, which is transmitted to the servo controller 7.

[0104] f: The servo controller 7 controls the servo group 8 by sending PWM waves according to the acquired control quantity, thereby realizing coordinated motion control toward the target.

[0105] The fuzzy PID control algorithm used has a faster response, smaller overshoot, and shorter adjustment time than traditional PID control, which improves the superiority of the controller. Figure 7 .

[0106] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

Claims

1. A binocular different-direction motion control method for a binocular robot, characterized in that: include: When entering the search mode, the two image sensors are controlled to move in opposite directions and the images captured by the two image sensors are continuously acquired to determine whether the captured images contain the target object. If so, the target tracking mode is entered; When entering the target tracking mode, if only one image sensor captures an image containing the target object, the image sensor that has not detected the target object is controlled to rotate in the same direction as the image sensor that has detected the target object until the images captured by both image sensors contain the target object; When entering the target tracking mode, if the images captured by both image sensors contain the target object, the deviation between the target object's position and the center of the camera's field of view, as well as the deviation increment between the previous and next frames, are sequentially fuzzified, fuzzy inference is performed according to the fuzzy rule table, and the fuzzification is performed to obtain the control amount of the two image sensors in the horizontal and vertical directions. The obtained control amount of the two image sensors in the horizontal and vertical directions is then distributed to each servo to obtain the control amount of each servo; The process of generating the control amounts of the two image sensors in the horizontal and vertical directions includes: Perform fuzzy processing on the input deviation and deviation increment, select 7 split points, denoted as NB, NM, NS, ZO, PS, PM, PB, and evenly split the deviation and deviation increment domain into eight intervals. At the same time, these 7 split points are used as fuzzy subsets. The linear membership function is used to calculate the membership of the lateral quantities of the deviation and the deviation increment; Perform fuzzy reasoning according to the fuzzy rule table, defuzzify the membership results, and obtain the increment ΔU of the lateral component according to the centroid method. w and the increment of the longitudinal component ΔU h ; Introducing coefficient λ to ΔU w and ΔU h Perform scaling and update the horizontal and vertical components of the PID output parameters; The updated PID output parameters are input into the PID controller to obtain the control quantities for the two image sensors in the horizontal and vertical directions.

2. The binocular different-direction motion control method of a binocular robot according to claim 1, characterized in that: The membership values ​​of the lateral quantity of the deviation and the deviation increment are both two, and the membership value of the lateral quantity of the deviation is C i 、C i+1 , for C i 、C i+1 The membership degrees are: (E w -C i ) / (C i+1 -C i )、(C i+1 -E w ) / (C i+1 -C i ) Where: E w is the lateral amount of deviation.

3. The binocular different-direction motion control method of a binocular robot according to claim 1, characterized in that: ΔU w The mathematical expression is: Among them: F i For U w Fuzzy subset, M i For U w F i The degree of membership.

4. The binocular different-direction motion control method of a binocular robot according to claim 1, characterized in that: The updating process of the lateral component of the PID output parameter is: U w (n)=U w (n-1)+λΔU w Among them: U w (n) is the updated lateral component, U w (n-1) is the horizontal component before updating.

5. The binocular different-direction motion control method of a binocular robot according to claim 1, characterized in that: The generation process of the control quantity of each servo includes: For the space directly in front of the fuselage, a two-dimensional rectangular coordinate system is established on a horizontal plane passing through the center of the camera and a vertical plane passing through the fuselage's central axis and perpendicular to the fuselage. The coordinate system is fixed to the fuselage, dividing the field of view into several areas. Among them, two image sensors are installed on the fuselage; Determine the area and position of the target object within the horizontal field of view; Assign control weights to the neck and eyes based on the target object's location. The eyes are the servos that control the two image sensors, and the neck is the servo that controls the fuselage. According to the assigned control weights, based on the obtained control amounts for the left and right eyes in the horizontal and vertical directions, they are distributed to each servo to obtain the control amount of each servo.

6. The binocular different-direction motion control method of a binocular robot according to claim 5, characterized in that: There are six steering gears in total, and the control quantity of each steering gear is as follows: Where: D out is the vector of all servo control quantities, D out1 is the control quantity of servo 1, D out2 is the control quantity of the second servo, D out3 is the control quantity of the servo 3, D out4 is the control quantity of the servo 4, D out5 is the control quantity of the servo 5, D out6 is the control quantity of servo six, C out_leftw is the horizontal control amount of the left image sensor, C out_lefth is the vertical control amount of the left image sensor, W neckw Control weight for the horizontal direction of the neck, C out_righth is the vertical control amount of the right image sensor, C out_rightw is the horizontal control amount of the right image sensor, W neckh is the control weight of the vertical direction of the neck, W lefth is the vertical control weight of the left image sensor, W leftw is the horizontal control weight of the left image sensor, W righth is the vertical control weight of the right image sensor, W rightw is the horizontal control weight of the right image sensor.

7. The binocular different-direction motion control method of a binocular robot according to claim 5, characterized in that: When only one image sensor captures an image containing the target object, the control weight of the neck is 0.

8. A target search device based on a binocular skew motion robot, comprising a memory, a processor, and a program stored in the memory, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 7 is implemented.

9. A storage medium having a program stored thereon, characterized in that: When the program is executed, the method according to any one of claims 1 to 7 is implemented.

Images