A control system and method for attitude-controlled experimental water tank vehicle

Abstract

The present invention discloses a posture-controlled experimental pool aerial vehicle control system and method, which belongs to the field of aerial vehicle control. The system includes an image acquisition module, a host computer, and an aerial vehicle system, wherein the image acquisition module and the aerial vehicle system are both connected to the host computer. The image acquisition module acquires image information of user skeleton nodes. The host computer converts the received image information into user skeleton node positions, and performs logical judgment with the user action skeleton node positions of preset control instructions, and issues instructions. The aerial vehicle system receives instructions issued by the host computer and controls the movement of the aerial vehicle. The user actions are used and controlled by acquiring user skeleton nodes and judging in real time through system preset logical judgment and support vector machine. The system can effectively enable multiple users to control the aerial vehicle in real time at different positions. The system can detect up to 6 controllers at the same time, and the algorithm has a recognition rate of 83% for user-defined actions on the test set.

Description

Technical Field

[0001] The invention belongs to the field of aerial vehicle control, and in particular relates to a control system and method for an aerial vehicle in an experimental water pool with attitude control. Background Art

[0002] In colleges and universities related to ocean science, experimental pools and anechoic pools are indispensable basic experimental facilities. When using a pool for testing, especially a large experimental pool, a pool trolley is a necessary auxiliary equipment, responsible for the placement and movement of personnel and equipment in the experimental area.

[0003] The pool trolley is a planar motion machine structure that can act on a vertical surface and can move horizontally. This technology was developed by A. Goodman in the United States in 1952. In 1989, M. Mackay and R. Walkes further developed it and proposed a mechanical device that can move in six degrees of freedom. There is little research on the pool trolley in China, and most of the research focuses on the improvement of the mechanical structure of the trolley and the exploration of traditional control. Wang Yang and others from Harbin Engineering University improved the traditional PID and used fuzzy reasoning to adjust the PID parameters to improve the robustness of the system. Patents CN110817686A and CN210286498U improve the mechanical structure of the traditional pool trolley and further improve the control accuracy by selecting the gear rack meshing walking and high inertia servo motor drive.

[0004] The existing tank drone system can only realize the traditional basic functions, including transporting personnel and equipment, and placing underwater equipment. However, when there are insufficient experimental personnel, the experiment is more complicated and requires multiple sets of repeated operations, the control of the tank drone becomes inefficient and repetitive. In addition, most drone control platforms are single, so personnel must control the drone in real time at the control platform, which wastes personnel and slows down the pace of the experiment. Summary of the invention

[0005] Technical issues to be solved:

[0006] In order to avoid the shortcomings of the prior art, the present invention provides a posture-controlled experimental pool aerial vehicle control system, which can collect the controller's skeleton nodes through the Kinect sensor, and compare the collected skeleton nodes with the custom node logic to perform real-time control. The user control area is greatly increased, the simplicity of user control is greatly improved, and the recording mode can make specific repeated movements without repeated input, which greatly optimizes the traditional aerial vehicle control mode. The inefficiency problem that a person must be on the platform for real-time control during traditional aerial vehicle control is solved.

[0007] The technical solution of the present invention is: a posture-controlled experimental pool vehicle control system, characterized in that: it includes an image acquisition module, a host computer, and a vehicle system, and the image acquisition module and the vehicle system are both connected to the host computer;

[0008] Collect image information of the user's skeleton nodes through the image acquisition module;

[0009] The received image information is converted into the user's bone node position through the host computer, and logical judgment is performed with the user's action bone node position of the preset control instruction, and an instruction is issued;

[0010] The vehicle system receives instructions from the host computer and controls the movement of the vehicle.

[0011] A further technical solution of the present invention is: the aerial vehicle system communicates with the host computer PC through a serial port.

[0012] A further technical solution of the present invention is: the image acquisition module is a Kinect sensor, which communicates with a host computer PC through an adapter.

[0013] A further technical solution of the present invention is: the host computer includes a PC, an SDK and a custom skeleton node processing algorithm, which is connected to the sensor and the vehicle system through the PC, and converts the image information into the user's skeleton node position coordinates through the SDK and the custom skeleton node processing algorithm.

[0014] A control method for a control system of an experimental water tank vehicle with attitude control, characterized in that the specific steps are as follows:

[0015] First, the host computer collects the image information of the user's skeletal nodes through the image acquisition module, determines whether there are users and the number of users, and analyzes the user's skeletal node parameters;

[0016] Then, the primary user is obtained through judgment and the control instruction of the primary user is identified;

[0017] Finally, the control instructions are transmitted to the aerial vehicle system through the serial port to control the aerial vehicle movement.

[0018] A further technical solution of the present invention is: the image acquisition module is a Kinect sensor; after collecting 20 bone nodes of the user through the Kinect sensor, logical judgment is performed on the collected user bone nodes, and when the user triggers the left wrist bone node to be higher than the head node, it is judged as raising the left hand high; when the right wrist bone node is triggered to be higher than the head node, it is judged as raising the right hand high; when the left wrist bone node is triggered to be higher than the shoulder node, it is judged as raising the left hand low; when the right wrist node is triggered to be higher than the shoulder node, it is judged as raising the right hand low; when both hands are raised at the same time, it is judged as raising both hands high or raising both hands low;

[0019] If the left wrist node moves to the right side of the right wrist node, it is determined to be pointing left; if the right wrist node moves to the left side of the left wrist, it is determined to be pointing right.

[0020] A further technical solution of the present invention is: the main user's judgment is that after the user turns on the system and raises his hands above his head, the system detects and determines that he is the main user. If the main user raises his hands above his head again, the system will turn off the somatosensory control function; another controller can continue to use it by raising his hands above his head.

[0021] A further technical solution of the present invention is: the recognition of the main user control instruction, the user's low-raising left hand indication action is used to enter the fine-tuning control mode, and the vehicle moves at a low speed, the user's low-raising right hand indication action is used to enter the normal control mode, and the vehicle moves at a medium speed; the user triggers the left and right indication action system to move left and right in the X-axis direction, and the user triggers the high-raising left and right hands indication action system to move left and right in the Y-axis direction.

[0022] A further technical solution of the present invention is: the user can formulate actions by self-setting joint node logic; the operator customizes related actions, and after shooting related actions as samples, the system will use soft-margin support vector machine to process the skeleton node coordinates.

[0023] Beneficial Effects

[0024] The beneficial effects of the present invention are as follows: the control system of the experimental water pool vehicle with posture control of the present invention realizes the use of user actions and controls the water pool vehicle by collecting user skeleton nodes and judging in real time through system preset logic discrimination and support vector machine. The system can effectively enable multiple users to control the vehicle in real time at different positions. The system can detect up to 6 controllers at the same time, and the algorithm has a recognition rate of 83% for user-defined actions on the test set.

[0025] For traditional aerial vehicle control, this control method reduces the user's control time by about 30 seconds per time, and the overall experimental time saving will increase with the complexity of the user's repeated actions. For the underwater acoustic signal propagation attenuation experiment, the single operation time is saved by about 1 minute, and one aerial vehicle control personnel is saved. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is the basic system framework diagram of the present invention.

[0027] Figure 2 This is a system signal transmission flow chart of the present invention.

[0028] Figure 3 It is the work flow chart of the present invention.

[0029] Figure 4 This is a display diagram of the identification interface of the present invention. DETAILED DESCRIPTION

[0030] The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0031] Reference Figure 1 , 2 As shown, Figure 1 The following is the basic system framework diagram of this system, which shows the basic components and framework structure of the entire system, and also shows the connection between each module; Figure 2 This is the signal transmission flow chart of this system, showing the signal type and signal flow mode of each link. This example is an experimental pool aerial vehicle control system. The system includes Kinect v2 as the skeleton node acquisition part, the computer as the host part, and the aerial vehicle system communicates with the computer through serial port. The Kinect sensor collects image information and transmits the image information to the host PC for processing. The host computer converts the image information into the user's skeleton node coordinates using Kinect SDK2 and makes a judgment. At the same time, the user-defined action is classified and learned using a support vector machine to obtain a model, which the user can use further. The main user is determined through judgment and the control command of the main user is identified. The control command is transmitted to the traditional aerial vehicle system through the serial port for control.

[0032] The system can be used by multiple users to perform gesture control on the vehicle, on the ground or in other control areas. When the user wants to control the vehicle, the user raises his hands. After the system detects, the user can make relevant gestures to control it. After the user control ends, the user raises his hands again to exit the gesture control mode and another user raises his hands to control it. When there is no user control, the vehicle can also be controlled using the traditional control mode. After entering the control mode, the vehicle can detect the user's bone node logic in real time, and the system will give priority to judging according to the user's customized actions. The user can also use the preset action to exit the customized action mode.

[0033] In the present embodiment, the traditional aerial vehicle part uses the traditional aerial vehicle mechanical part, so that the system can adapt to any aerial vehicle and can be conveniently transplanted to any aerial vehicle for user use, and communicates with the host computer PC end through the serial port, so that the system is convenient to transplant. The Kinect sensor is connected to the USB port of the PC end through the official converter, is responsible for collecting the image information of the user, and transmits the data to the computer for processing. The host computer PC end uses KinectSDK to obtain the user's skeleton node position by receiving the image information transmitted by the Kinect sensor, and performs logical judgment with the user's action skeleton node position of the preset control instruction, and the user can customize the skeleton node of the relevant control action at any time. The host computer will send the control instruction to the aerial vehicle for processing through the serial port.

[0034] Reference Figure 3 , 4 As shown, Figure 3 The system workflow diagram described here shows the entire system's logical working mode and the status of each system instruction. Figure 4 The following is a display diagram of the recognition interface of this system, which shows the skeletal states of three users identified by the system in a certain situation and the related action information of triggering "sit down". The three LOGOs in the figure indicate that three users are identified to be using the system. The preset action is "sit down". At this time, two users trigger this action. It can be seen that two LOGOs change and one LOGO remains unchanged. The right side of the figure shows the skeletal nodes. After the host computer collects the user image, it first determines whether there is a user, how many users there are, and analyzes the skeletal node parameters of the user. After collecting 20 skeletal nodes of the user, logical judgment is performed on the collected user skeletal nodes. If the user triggers the left wrist skeletal node higher than the head node, it is judged as raising the left hand high, if the right wrist skeletal node is higher than the head node, it is judged as raising the right hand high, if the left wrist skeletal node is higher than the shoulder node, it is judged as raising the left hand low, and if the right wrist skeletal node is higher than the shoulder node, it is judged as raising the right hand low. Similarly, raising both hands at the same time can also be judged as raising both hands high or raising both hands low. If the left wrist node moves to the right side of the right wrist node, it is determined to indicate the left, and if the right wrist node moves to the left side of the left wrist, it is determined to indicate the right. After the user turns on the system, the system will detect and determine that he is the main user. If the main user raises his head again, the system will turn off the somatosensory control function, and another controller can continue to use it by raising both hands. The user's left hand indication action is used to enter the fine-tuning control mode, and the drone moves at a low speed. The user's right hand indication action is used to enter the normal control mode, and the drone moves at a medium speed. The user triggers the left and right indication action system to move left and right in the X-axis direction, and the user triggers the left and right hand indication action system to move left and right in the Y-axis direction. At the same time, the upper computer will present the current somatosensory user information through the display screen. Users can also set their own joint node logic to formulate actions to map related actions to instructions that need to be repeatedly executed, so that the drone can perform more complex and repetitive operations with only one action.

[0035] In the self-set joint mode, the operator customizes the relevant actions. After shooting the relevant actions as samples, the system will use the soft margin support vector machine to process the bone node coordinates. The relevant training set and test set are formulated by the user's labeled bone node coordinates using K-fold cross validation. The user input data is:

[0036] D k ={(x1, x1 * ), (x2, x2 *),…,(x 20 , x 20 * )},y i = ±1

[0037] Among them, (x i ,y i ) is the coordinate of the i-th bone node when the user completes the action numbered k. The hyperplane divides the skeleton nodes of each action. The distance from the coordinate to the hyperplane is In order to use a hyperplane to divide each sample, it is only necessary to maximize the distance between the two heterogeneous samples closest to the hyperplane and the hyperplane.

[0038]

[0039]

[0040] Add Lagrange multiplier α i >0, its dual problem can be transformed into:

[0041]

[0042] sty i α i =0, i=1, 2, ...

[0043] α i ≥0

[0044] After solving, the classification model can be obtained:

[0045]

[0046] When the user uses gesture control to control the drone, first use the preset action to control the drone to the specified position, and then use the defined action to conduct control experiments on the drone system.

[0047] The signal flow of the entire system is that the Kinect sensor first collects image information and transmits the signal to the host PC for processing. The host PC converts the image information into the user's skeletal node coordinates using the Kinect SDK, uses the judgment algorithm and makes a judgment, and obtains the main user through the judgment and identifies the main user's control instructions. The control instructions are transmitted to the traditional aerial vehicle system through the serial port for control.

[0048] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention without departing from the principles and intent of the present invention.

Claims

1. A control method for a posture-controlled experimental tank vehicle control system, characterized in that The specific steps are as follows: First, the host computer collects the image information of the user's skeletal nodes through the image acquisition module, determines whether there are users and the number of users, and analyzes the user's skeletal node parameters; Then, the primary user is obtained through judgment and the control instruction of the primary user is identified; Finally, the control instructions are transmitted to the vehicle system through the serial port to control the vehicle motion; The image acquisition module is a Kinect sensor; after collecting 20 bone nodes of the user through the Kinect sensor, logical judgment is performed on the collected user bone nodes. If the user triggers the left wrist bone node to be higher than the head node, it is judged as raising the left hand high; if the user triggers the right wrist bone node to be higher than the head node, it is judged as raising the right hand high; if the left wrist bone node is higher than the shoulder node, it is judged as raising the left hand low; if the right wrist node is higher than the shoulder node, it is judged as raising the right hand low; if both hands are raised at the same time, it is judged as raising both hands high or low; if the left wrist node moves to the right side of the right wrist node, it is judged as indicating to the left; if the right wrist node moves to the left side of the left wrist, it is judged as indicating to the right; The decision of the main user: after the user turns on the system, the system will detect that the user is the main user by raising his hands above his head. If the main user raises his hands above his head again, the system will turn off the somatosensory control function; another controller can continue to use the system by raising his hands above his head; The recognition of the main user control command, the user raises the left hand to indicate the action to enter the fine-tuning control mode, the vehicle moves at a low speed, the user raises the right hand to indicate the action to enter the normal control mode, the vehicle moves at a medium speed; the user triggers the left and right indication action system to move left and right in the X-axis direction, the user triggers the left and right hand indication action system to move left and right in the Y-axis direction; The user can formulate actions by setting the joint node logic by himself; the operator customizes the relevant actions, and after shooting the relevant actions as samples, the system will use the soft margin support vector machine to process the skeleton node coordinates.

2. A posture-controlled experimental tank vehicle control system, used to implement the control method of claim 1; characterized in that: It includes an image acquisition module, a host computer, and a vehicle-mounted vehicle system, wherein the image acquisition module and the vehicle-mounted vehicle system are both connected to the host computer; Collect image information of the user's skeleton nodes through the image acquisition module; The received image information is converted into the user's bone node position through the host computer, and logical judgment is performed with the user's action bone node position of the preset control instruction, and an instruction is issued; The vehicle system receives instructions from the host computer and controls the movement of the vehicle.

3. According to the attitude-controlled experimental water tank vehicle control system of claim 2, it is characterized by: The aerial vehicle system communicates with the host computer PC through the serial port.

4. The control system for an experimental water tank vehicle with attitude control according to claim 2, characterized in that: The image acquisition module is a Kinect sensor, which communicates with the host computer PC through an adapter.

5. The control system for an experimental water tank vehicle with attitude control according to claim 2, characterized in that: The host computer includes a PC, SDK and a custom skeleton node processing algorithm, which is connected to the sensor and the vehicle system through the PC, and converts image information into user skeleton node position coordinates through the SDK and the custom skeleton node processing algorithm.

Images