Multi-interface multi-link unmanned aerial vehicle autopilot and control method thereof

Abstract

The invention relates to the technical field of unmanned aerial vehicles, in particular to a multi-interface and multi-link unmanned aerial vehicle autopilot and a control method thereof.The multi-interface module is used for controlling dynamic access switching of an unmanned aerial vehicle and a multi-link interface, a stepping motor drives a lead screw to drive a sliding block to move axially, a target interface unit is selected from a plurality of interface units, and the target interface unit is connected with the unmanned aerial vehicle; the brushless motor is controlled to drive the sliding sleeve to rotate in the butt joint process, so that the target interface unit establishes directional electric connection with a CPU chip in the chip bin through an electric slip ring type contact structure, and electromagnetic crosstalk of multi-interface parallel transmission is inhibited; the chip bin is used for guiding airflow generated by flying of the unmanned aerial vehicle to form a directional air film through the wedge-shaped gap and the linear airflow groove in the edge of the water drop, and the contact thermal resistance between the chip bin and the interface unit is reduced. Through cooperative work of the multi-interface module and the chip bin, efficient dynamic access switching and airflow guiding heat dissipation of the unmanned aerial vehicle and the multi-link interface are achieved, and the heat dissipation efficiency and operation stability of the system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles, and particularly to an unmanned aerial vehicle autopilot with multiple interfaces and multiple links and a control method thereof. Background Art

[0002] With the continuous improvement of the integration level of civilian unmanned aerial vehicle systems, the interaction requirements between their functional modules and ground systems have become increasingly complex. The requirements for multi-device collaborative communication and efficient heat dissipation have become the key technical bottlenecks restricting system performance. In the existing unmanned aerial vehicle system architecture, the communication link between the flight control system and the ground station has been difficult to meet the requirements of collaborative work of airborne devices in multi-task scenarios. On the one hand, high-precision navigation devices such as differential GPS mobile stations need to receive navigation message correction data uploaded by ground base stations in real time, while mission payload devices need to interact with ground control terminals for high-frequency status instructions. The data interface types and protocol standards between different devices are significantly different, resulting in the need to configure multiple independent communication modules at the ground end, increasing the complexity of system integration and operation difficulty.

[0003] The traditional unmanned aerial vehicle interface design has significant limitations. Taking the differential GPS system as an example, the correction data is usually transmitted between the ground base station and the airborne mobile station through a point-to-point radio frequency link. However, limited by the transmission power and antenna directivity, communication interruptions are likely to occur in complex terrain environments. Mission payload devices mostly use independent data links, and their interface specifications vary from device manufacturers, resulting in the need to configure multiple physical interface conversion devices at the ground control terminal, which not only increases the system weight but also reduces the emergency response speed. When the unmanned aerial vehicle performs multi-task collaborative operations, the data synchronization requirements between different devices make it necessary to configure multiple independent receivers at the ground end, exacerbating the risk of electromagnetic interference and device power consumption. At the same time, in terms of heat dissipation design, the existing autopilot system faces severe thermal management challenges. Currently, most solutions adopt a combination of passive heat dissipation and natural convection, and the heat of the CPU chip is conducted to the casing through an aluminum alloy heat sink. However, with the increase in the complexity of navigation algorithms and real-time processing requirements, the chip power consumption density has increased exponentially, and the traditional heat dissipation structure has been difficult to control the temperature rise amplitude.

[0004] In summary, the limitations of existing interface and heat dissipation technologies seriously restrict the development of unmanned aerial vehicle systems. These technical bottlenecks not only reduce system reliability but also increase operation and maintenance costs, becoming the key obstacles restricting the expansion of unmanned aerial vehicles to more extensive application fields. Therefore, developing an autopilot system with multi-interface intelligent management and high-integration heat dissipation capabilities has become the core technical requirement for improving the overall performance of unmanned aerial vehicles. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides an unmanned aerial vehicle autopilot with multiple interfaces and multiple links and a control method thereof.

[0006] In a first aspect, the present invention provides a multi-interface multi-link unmanned aerial vehicle autopilot, which includes a multi-interface module and a chip bin; wherein, the multi-interface module includes a number of interface units, a brushless motor, a lead screw, a slider, a sliding sleeve, and a stepper motor; The multi-interface module is used to control the dynamic access switching of the unmanned aerial vehicle with the multi-link interface, and drive the lead screw through the stepper motor to drive the slider to axially move, so as to select a target interface unit from multiple interface units, and control the brushless motor to drive the sliding sleeve to rotate during the docking process, so that the target interface unit establishes a directional electrical connection with the CPU chip in the chip bin through an electro-slip ring type contact structure, suppressing the electromagnetic crosstalk of multi-interface parallel transmission; The chip bin is used to form a directional air film through the wedge-shaped gap and the linear airflow groove at the edge of the water droplet to guide the airflow generated when the unmanned aerial vehicle flies, reducing the contact thermal resistance between the chip bin and the interface unit.

[0007] In a further embodiment, each of the interface units includes an integrally formed metal shell and a rotating sleeve, and sliding grooves are provided on the outer circumference of the sliding sleeve and the rotating sleeve along the length direction; The rotating sleeve of each interface unit is rotatably connected to the outer circumference of the sliding sleeve through the sliding groove along the length direction of the sliding sleeve.

[0008] In a further embodiment, the output shaft of the brushless motor is fixedly connected to the sliding sleeve through the housing of the stepper motor, and both ends of the sliding sleeve are rotatably connected to the top of the chip bin through a pair of sliding sleeve mounting seats, and the sliding sleeve rotates around the axis of the sliding sleeve mounting seat; The brushless motor is used to drive the sliding sleeve to rotate by driving the stepper motor, and then drive the interface unit connected thereto to rotate to the working position.

[0009] In a further embodiment, electro-slip ring type contact structures are provided on the outer wall of the sliding sleeve and the inner wall of the rotating sleeve; The electro-slip ring type contact structure on the outer wall of the sliding sleeve is connected to the CPU chip in the chip bin through a data line, and when the electro-slip ring type contact structure on the inner wall of the rotating sleeve contacts the electro-slip ring type contact structure on the outer wall of the sliding sleeve when the selected target interface unit rotates to a preset angle, the selected target interface unit forms a directional electrical connection with the CPU chip in the chip bin in a rotating state, suppressing the electromagnetic crosstalk when multiple interface units work in parallel.

[0010] In a further embodiment, the output shaft of the stepper motor is fixedly connected to the lead screw, and the lead screw is coaxially arranged in the sliding sleeve and is threadedly connected to the slider; The stepper motor drives the lead screw to drive the slider to axially move within the sliding sleeve. During the process of the slider sliding along the sliding sleeve, the slider cooperates with the sliding grooves of the rotating sleeves of the respective interface units in sequence, so that the slider locks or unlocks different interface units in sequence, thereby controlling the dynamic switching connection of the multi-link interfaces.

[0011] In a further embodiment, the chip bin includes a rectangular box-shaped bin housing, and a CPU chip and a battery are encapsulated in the bin housing; a water-droplet-edge linear air flow groove penetrating the front and rear ends is provided on the upper surface of the bin housing, and a wedge-shaped gap is formed between the front end of the upper surface of the bin housing and the same side of the interface unit; The water-droplet-edge linear air flow groove is used to guide the flight air flow of the drone to flow unidirectionally in the front-rear direction through the wedge-shaped gap between the chip bin and the interface unit during the flight of the drone, generate a directional dynamic pressure air film through the dynamic pressure effect, reduce the accumulation of frictional heat on the contact surface between the chip bin and the interface unit, and thereby offset the temperature rise effect of the CPU chip under high-load conditions.

[0012] In a further embodiment, each interface unit is respectively configured as at least four of a network interface, a USB interface, an I / O interface, an HDMI interface, a debugging interface, and a power interface.

[0013] In a second aspect, the present invention provides a control method for a multi-interface multi-link drone autopilot, and the control method includes the following steps: In response to a drone docking request, determine a target interface unit according to the drone docking request, and control the slider to axially move to the target interface unit for locking; Drive the slider to drive the target interface unit to rotate from a horizontal storage position to a vertical working position, and collect a target interface image of the target interface unit; Preprocess the target interface image to obtain a preprocessed interface image, and extract the target interface features of the target interface unit from the preprocessed interface image by using a convolutional neural network; the convolutional neural network includes hybrid pooling, pyramid pooling, and a residual network structure; Match and verify the target interface features with the standard interface unit features. When the match verification fails, control the brushless motor to finely adjust the rotation angle of the target interface unit until the match verification passes; When the match verification passes, drive the target interface unit to establish an external physical connection with the chip bin and start data transmission; After the data transmission is completed, disconnect the external physical connection and control the brushless motor to drive the target interface unit to reset from the vertical working position to the horizontal storage position, and prepare to execute the next interface switch.

[0014] In a further embodiment, the target interface features include a target interface shape feature and a target interface position feature; the step of matching and verifying the target interface features with the standard interface unit features includes: Perform shape matching between the target interface shape feature and a preset standard interface unit shape feature, calculate the overlapping area ratio between the target interface shape feature and the preset standard interface unit shape feature based on the Dice similarity coefficient, and obtain a shape matching degree; Use a pre-constructed shape loss function to quantify the shape matching error; the shape loss function consists of a Dice similarity coefficient loss term and a shape deviation penalty term; If the shape matching error is lower than a preset shape error threshold range, it is determined that the matching verification of the target interface shape feature passes; Detect the key corner positions of the target interface unit according to the target interface position feature, and extract the corner coordinates from the key corner positions; Introduce a position deviation penalty term and an Euclidean distance loss term, construct a position loss function, and use the position loss function to quantify the position loss value corresponding to the coordinate deviation between the corner coordinates and the standard interface unit position feature; When the position loss value meets a preset position error threshold range, it is determined that the matching verification of the target interface position feature passes.

[0015] In a further embodiment, the control method further includes: During data transmission, detect the local temperature of the chip bin in real time, and when it is detected that the local temperature of the chip bin exceeds a preset chip bin temperature threshold range, actively control the brushless motor to drive the target interface unit to rotate, increase the wedge gap separation angle between the target interface unit and the chip bin, and constrain the unidirectional flow of the airflow generated during the flight of the drone through the linear airflow groove at the edge of the water droplet until the local temperature of the chip bin drops back to within the preset chip bin temperature threshold range.

[0016] The present invention provides a multi-interface multi-link unmanned aerial vehicle autopilot and its control method. The unmanned aerial vehicle autopilot includes a multi-interface module and a chip bin. The multi-interface module is used to control the dynamic access switching between the unmanned aerial vehicle and the multi-link interface, and drives the lead screw through the stepper motor to drive the slider to axially move, so as to select a target interface unit from multiple interface units, and controls the brushless motor to drive the sliding sleeve to rotate during the docking process, so that the target interface unit establishes a directional electrical connection with the CPU chip in the chip bin through an electro-slip ring type contact structure, suppressing the electromagnetic crosstalk of multi-interface parallel transmission. The chip bin is used to form a directional air film by guiding the airflow generated during the flight of the unmanned aerial vehicle through the wedge-shaped gap and the linear airflow groove at the edge of the water droplet, reducing the contact thermal resistance between the chip bin and the interface unit. Compared with the prior art, through the collaborative work of the multi-interface module and the chip bin, the unmanned aerial vehicle autopilot realizes the efficient dynamic access switching between the unmanned aerial vehicle and the multi-link interface, and at the same time combines the wedge-shaped gap and the linear airflow groove at the edge of the water droplet to guide the airflow to form a directional air film, reducing the contact thermal resistance between the chip bin and the interface unit, improving the heat dissipation efficiency and operation stability of the system, and meeting the high-performance communication and navigation control requirements of the unmanned aerial vehicle in complex flight environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a schematic structural diagram of a multi-interface multi-link unmanned aerial vehicle autopilot provided by an embodiment of the present invention; Figure 2 is a schematic cross-sectional structural diagram of the unmanned aerial vehicle autopilot provided by an embodiment of the present invention; Figure 3 is a schematic overall structural diagram of the unmanned aerial vehicle autopilot provided by an embodiment of the present invention; Figure 4 is a schematic structural diagram of the chip bin provided by an embodiment of the present invention; Figure 5 is a schematic flow diagram of the control method of the multi-interface multi-link unmanned aerial vehicle autopilot provided by an embodiment of the present invention; Figure 6 is a process block diagram of the control method of the unmanned aerial vehicle autopilot provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The following specifically illustrates the embodiments of the present invention in conjunction with the drawings. The presentation of the embodiments is only for the purpose of illustration and should not be construed as a limitation of the present invention. The drawings are only for reference and illustration, and do not constitute a limitation on the scope of patent protection of the present invention, because many changes can be made to the present invention without departing from the spirit and scope of the present invention.

[0019] Referring to Figure 1 , an embodiment of the present invention provides a multi-interface multi-link unmanned aerial vehicle autopilot, as Figure 1As shown in the figure, the UAV autopilot includes a multi-interface module 1 and a chip bin 2. The UAV autopilot is installed on the UAV 3 and is used to control and navigate the UAV 3. In this embodiment, as Figure 2 shown, the multi-interface module 2 includes a number of interface units 101, a brushless motor 11, a lead screw 12, a slider 13, a sliding sleeve 14, and a stepper motor 15. The chip bin serves as the computing and energy core of the autopilot. The chip bin includes a rectangular box-shaped bin housing, and key components such as a CPU chip and a battery are loaded inside the bin housing, ensuring the stable operation and long endurance of the UAV autopilot in a complex flight environment.

[0020] In some embodiments, the multi-interface module is used to control the dynamic access switching of the UAV and the multi-link interface, and drives the lead screw through the stepper motor to drive the slider to axially move, so as to select a target interface unit from multiple interface units, and controls the brushless motor to drive the sliding sleeve to rotate during docking, so that the target interface unit establishes a directional electrical connection with the CPU chip in the chip bin through an electro-slip ring type contact structure, suppressing the electromagnetic crosstalk of multi-interface parallel transmission.

[0021] As Figure 3 shown, the multi-interface module 1 serves as the information interaction center of the UAV autopilot. The multi-interface module integrates multiple interface units 101, and each interface unit 101 corresponds to a different data interface type respectively to adapt to different data connection requirements. In some embodiments, each interface unit is respectively configured as at least four of a network interface, a USB interface, an I / O interface, an HDMI interface, a debugging interface, and a power interface to meet the diverse data transmission and control instruction reception requirements.

[0022] In this embodiment, each interface unit 101 includes an integrally formed metal shell 1011 and a rotating sleeve 1012. The sliding sleeve 14 is a key component connecting the multi-interface module 1 and the chip bin 2. The sliding sleeve 14 is installed on the top of the chip bin 2. A pair of sliding sleeve mounting seats 21 are fixedly installed on the top of the chip bin 2 to fixedly support the sliding sleeve 14 through the sliding sleeve mounting seats 21 to ensure its stability without shaking. The front and rear ends of the sliding sleeve 14 are respectively rotatably connected to the two sliding sleeve mounting seats 21, so that the sliding sleeve 14 can rotate flexibly within a certain range to adapt to different connection requirements, that is, the two ends of the sliding sleeve are rotatably connected to the top of the chip bin through a pair of sliding sleeve mounting seats, and the sliding sleeve rotates around the axis of the sliding sleeve mounting seat. At the same time, the rotating sleeves 1012 of each interface unit 101 are rotatably connected to the outer circumference of the sliding sleeve along the length direction of the sliding sleeve through the sliding groove. Such a layout not only saves space but also enables the interface unit 101 to flexibly adjust its position and angle as needed.

[0023] To further enhance the stability and reliability of the connection, sliding grooves 102 are provided on the outer peripheries of the sliding sleeve 14 and the rotating sleeve 1012 in the length direction. The output shaft of the stepping motor is fixedly connected to the lead screw. The lead screw is coaxially arranged inside the sliding sleeve and is threadedly connected to the slider. The stepping motor drives the lead screw to drive the slider to axially move inside the sliding sleeve. During the sliding of the slider 13 along the sliding sleeve 14, the slider 13 cooperates with the sliding grooves of the rotating sleeves of each interface unit in sequence. As the slider 13 slides, the slider 13 locks or unlocks the sliding sleeve 14 and different interface units 101 in sequence, thereby ensuring the stability and reliability of the interface unit 101 during the connection process, and at the same time realizing the control of the dynamic switching connection of multiple-link interfaces and improving the work efficiency.

[0024] In this embodiment, the brushless motor is used to drive the stepping motor to drive the sliding sleeve to rotate, and further drive the interface unit connected thereto to rotate to the working position; the output shaft of the brushless motor 11 is fixedly connected to the housing of the stepping motor 15, and the housing of the stepping motor 15 is fixedly connected to the sliding sleeve 14, so that when the brushless motor 11 starts, the output shaft of the brushless motor 11 can drive the sliding sleeve 14 to rotate, and further drive the interface unit 101 locked to the sliding sleeve 14 to rotate synchronously, realizing the flexible adjustment of the interface unit 101.

[0025] In this embodiment, a number of electrical slip ring type contact structures are provided on the outer wall of the sliding sleeve 14 and the inner wall of the rotating sleeve 1012. The electrical slip ring type contact structure on the outer wall of the sliding sleeve 14 is connected to the CPU chip in the chip bin 2 through a data line, ensuring stable signal transmission. In this embodiment, when the metal shell 1011 of the interface unit 101 is in contact with the bin shell 201 of the chip bin 2, the contacts on the outer wall of the sliding sleeve 14 and the contacts on the inner wall of the rotating sleeve 1012 are in a separated state. At this time, no electrical connection is formed between the interface unit 101 and the CPU chip. When the selected target interface unit 101 rotates to a preset angle, the electrical slip ring type contact structure on the outer wall of the sliding sleeve 14 and the electrical slip ring type contact structure on the inner wall of the rotating sleeve 1012 come into contact, thereby realizing the directional electrical connection between the selected target interface unit in the rotating state and the CPU chip, and suppressing the electromagnetic crosstalk when multiple interface units work in parallel. It should be noted that the electrical slip ring structures provided on the contacts on the outer wall of the sliding sleeve 14 and the inner wall of the rotating sleeve 1012 are similar to an electrical slip ring (Electrical Slip Ring). The electrical slip ring allows the mechanical structure to maintain a stable electrical connection while rotating infinitely at 360°. Therefore, through this design, this embodiment can not only realize the flexible rotation of the interface unit 101, but also ensure that the interface unit 101 can reliably transmit data and supply power to the CPU chip during the rotation process, providing a strong guarantee for the normal operation of the device.

[0026] In this embodiment, a wedge-shaped gap 202 is provided between the front end of the upper surface of the bin shell 201 and the front end of the lower surface of the metal shell 1011, so as to utilize the airflow generated by the drone 3 during flight to enhance the heat dissipation effect of the CPU chip. The water droplet edge linear air flow groove is used to guide the flight airflow of the drone through the wedge-shaped gap to flow unidirectionally in the front-rear direction between the chip bin and the interface unit during the flight of the drone, generating a directional dynamic pressure air film through the dynamic pressure effect, reducing the friction heat accumulation on the contact surface between the chip bin and the interface unit, and further offsetting the temperature rise effect of the CPU chip under high-load conditions. Figure 4 It is a schematic diagram of the chip bin structure.

[0027] Specifically, when the drone 3 is flying, it generates an air current. The air current enters the space between the chip bin 2 and the interface unit 101 through the wedge-shaped gap 202. Based on the hydrodynamic pressure effect, a dynamic air film is formed between the chip bin 2 and the interface unit 101. This air film not only plays an isolation role, but also reduces the contact surface thermal resistance and enhances the heat dissipation efficiency of the CPU chip. To further optimize the flow path of the air current, in this embodiment, an air current groove 203 is arranged on the upper surface of the bin housing 201. The air current groove 203 runs through the bin housing 201 in the front-rear direction. The air current groove 203 is composed of a number of air current groove units in the shape of the edge line of a water droplet connected to each other. The shape of each air current groove unit is the edge line shape of a water droplet with a larger front and a smaller rear. The setting of the air current groove 203 not only ensures the stable opening of the air film, but also can guide the air current to flow back and forth in a specific direction. The air current will flow into from the side where the wedge-shaped gap 202 is provided between the chip bin 2 and the interface unit 101, and then flow along the direction of the air current groove 203 to the other side, that is, guide the air current to flow unidirectionally from the side of the wedge-shaped gap 202 to the other side. This unidirectional flowing air current not only improves the heat dissipation efficiency, but also ensures the stability and controllability of the entire heat dissipation process. Therefore, in this embodiment, by arranging the wedge-shaped gap 202 and the air current groove 203, an efficient and stable CPU chip heat dissipation solution is realized, which not only improves the overall performance of the drone, but also provides a strong guarantee for its long-term stable operation.

[0028] In one embodiment, as Figure 5 shown, the embodiment of the present invention provides a control method for a multi-interface multi-link drone autopilot. The control method includes the following steps: S1. In response to a drone docking request, determine a target interface unit according to the drone docking request, and control the slider to axially move to the target interface unit for locking.

[0029] S2. Drive the slider to drive the target interface unit to rotate from the horizontal storage position to the vertical working position, and collect a target interface image of the target interface unit.

[0030] Specifically, as Figure 6 shown, in this embodiment, in response to a drone docking request, determine the interface type required for docking by the drone according to the communication protocol between the drone and the drone autopilot, and determine the target interface unit according to the interface type required for docking by the drone. Then control the stepper motor to drive the slider to move to the sliding groove position of the corresponding interface. The brushless motor rotates to drive the slider to further rotate, so that the selected target interface unit changes from the horizontal state to the vertical working position, ready for docking. At the same time, to ensure that the interface unit can be accurately and safely connected, the drone uses the camera carried by itself to capture the target interface image of the target interface unit to achieve the accurate connection of the interface unit.

[0031] S3. Preprocess the target interface image to obtain a preprocessed interface image, and extract the target interface features of the target interface unit from the preprocessed interface image by using a convolutional neural network; the convolutional neural network includes hybrid pooling, pyramid pooling, and a residual network structure.

[0032] In this embodiment, the acquired target interface image is transmitted to the CPU chip, and preprocessing operations are performed on the target interface image to obtain a preprocessed interface image. Among them, the preprocessing operations include image segmentation, filtering, enhancement, and denoising, etc., to improve the quality and stability of the image. During this process, this embodiment uses preprocessing tools such as median filtering, Gaussian filtering, morphological operations, binarization, or histogram equalization to effectively remove the noise in the image, enhance the target features, and ensure that the image quality meets the subsequent recognition requirements. Then, this embodiment extracts the target interface shape features and target interface position features of the target interface unit from the preprocessed interface image to form target interface features. Among them, this embodiment selects a convolutional neural network to extract features from the preprocessed image according to the task type and data characteristics. At the same time, to achieve more refined feature extraction, this embodiment uses an improved convolutional neural network structure. During the process of recognizing the shape of the target interface unit, this embodiment uses a multi-scale convolutional kernel structure. In the same convolutional layer, three different sizes of convolutional kernels, namely 3x3 small size, 5x5 medium size, and 7x7 large size, are used simultaneously to capture different scale features of the interface unit. Among them, the 3x3 small-size convolutional kernel can accurately capture the detailed features of the target interface unit, such as edges and corners, and enhance the sensitivity to small position changes; the 5x5 medium-size convolutional kernel is mainly used to capture the medium-scale features of the interface unit, such as larger contour changes and local shapes, and improve the extraction ability of contour and angle information; the 7x7 large-size convolutional kernel can capture the global features of the interface unit, reduce the influence of perspective transformation and scaling on shape recognition, and enhance the model's perception ability of the overall structure. The large-size convolutional kernel is usually placed in the relatively shallow layer of the network or the downsampled image layer to extract wide-range features and capture the overall structural features of the interface, so as to be able to capture both the local and global features of the interface unit, provide a stable recognition effect for the interface shape in different environments, and make correct judgments and processing.

[0033] The convolutional neural network structure in this embodiment realizes more refined feature extraction through multiple convolutional and pooling operations. In this embodiment, after the convolutional layer, a hybrid pooling strategy is introduced during the pooling process, achieving better adaptation to different-scale and detailed features. The hybrid pooling strategy alternately uses max pooling and average pooling in different convolutional layers. Max pooling is used to extract significant features and enhance the clarity of edges and corners; average pooling is used to smooth unimportant feature information and reduce noise interference. Then, in this embodiment, pyramid pooling is adopted in higher-level feature maps, dividing the higher-level feature maps into different-scale regions and performing pooling on each region separately to capture context information at different scales, enabling the model to perceive shape features at different scales, thereby enhancing the model's recognition ability for the overall and local features of the interface unit.

[0034] In addition, this embodiment introduces a residual network structure to ensure that more details and location information are retained during feature extraction. Each residual block not only includes convolutional + batch normalization + ReLU activation operations, but also adds a skip connection. Through the skip connection, more original details and location information are retained during feature propagation. In this embodiment, a bottleneck structure of 1x1, 3x3, 1x1 is adopted in the residual block. Among them, 1x1 convolution is used for dimensionality reduction and dimensionality increase to reduce the amount of computation; 3x3 convolution is used for feature extraction, retaining important feature details while reducing the amount of computation. The skip connection retains the original information during feature propagation, avoiding detail loss, further improving the recognition accuracy and reducing the model complexity.

[0035] In terms of position recognition, this embodiment introduces a hierarchical classification model based on the convolutional neural network structure. This hierarchical classification model determines the accessibility of the interface unit step by step. The first-layer model is used to determine whether the shape matching degree of the interface unit meets the access requirements, and the second-layer model verifies the position accuracy of the interface unit. This setting of distributed verification in the hierarchical classification model not only reduces the misjudgment rate but also significantly improves the real-time performance of recognition, thus achieving accurate recognition of the shape and position of the interface unit, providing a strong guarantee for the precise docking of devices such as UAV autopilots.

[0036] S4. Match and verify the target interface features with the standard interface unit features. When the match verification fails, control the brushless motor to finely adjust the rotation angle of the target interface unit until the match verification passes.

[0037] In some embodiments, the target interface features include target interface shape features and target interface position features; the step of matching and verifying the target interface features with the standard interface unit features includes: Perform shape matching between the target interface shape feature and the preset standard interface unit shape feature, calculate the overlapping area ratio of the target interface shape feature and the preset standard interface unit shape feature based on the Dice similarity coefficient, and obtain the shape matching degree; Quantify the shape matching error using a pre-constructed shape loss function; the shape loss function consists of a Dice similarity coefficient loss term and a shape deviation penalty term; If the shape matching error is lower than the preset shape error threshold range, it is determined that the matching verification of the target interface shape feature passes; Detect the key corner positions of the target interface unit according to the target interface position feature, and extract the corner coordinates from the key corner positions; Introduce a position deviation penalty term and an Euclidean distance loss term, construct a position loss function, and use the position loss function to quantify the position loss value corresponding to the coordinate deviation between the corner coordinates and the standard interface unit position feature; When the position loss value meets the preset position error threshold range, it is determined that the matching verification of the target interface position feature passes.

[0038] Specifically, in this embodiment, the target interface feature of the target interface unit is extracted from the preprocessed interface image. In order to extract the complete edge contour of the target interface unit, this embodiment uses edge detection algorithms such as Canny edge detection or Sobel operator to operate on the preprocessed interface image, extracts the complete edge contour of the target interface unit, and then matches the extracted complete edge contour of the target interface unit with the standard interface unit edge in the standard template. This embodiment uses the Hausdorff distance algorithm to calculate the difference between the actual edge and the template edge, so as to evaluate the matching degree. If the matching error is within the preset threshold range, it is determined that the edge verification passes; if the error is too large, it is determined that the target interface unit cannot be accessed. On this basis, this embodiment further detects the subtle features of the key parts on the edge, such as the subtle features of key parts such as specific bending points and concave-convex parts, and controls the errors of these key part subtle features within a stricter threshold to improve the recognition accuracy.

[0039] At the same time, this embodiment uses the Harris corner detection or Shi-Tomasi corner detection algorithm to identify the key corner positions of the target interface unit in the image. The corner detection algorithm accurately detects the main corner points of the target interface unit by analyzing the gradient of the image gray change, compares the detected corner positions with the corner positions in the standard template, and calculates the corner deviation. If the matching of the edge and the corner both meet the preset requirements, it is determined that the subtle feature verification of the target interface unit passes. At this time, the target interface unit is accurate and can be accessed. Otherwise, this embodiment will determine that the target interface unit cannot be accessed and readjust the position of the drone for re-verification.

[0040] Based on feature extraction and verification, in this embodiment, the extracted features are compared with known data, and the new data is classified into corresponding categories according to the pre-set classification rules. Then, according to system requirements, means such as experience, rules, or statistical methods are used to analyze and reason about the data. To improve the matching accuracy of shape and position, a dedicated comprehensive loss function is designed in this embodiment. This comprehensive loss function consists of two parts: a position loss function and a shape matching loss function, and penalty terms for position deviation and shape matching degree are added, so that the model pays more attention to the accuracy of shape and position during training. The comprehensive loss function is specifically expressed as: In the formula, is the comprehensive loss function; is the position loss function; is the shape matching loss function; and are both weight parameters, which are used to control the proportion of position and shape losses in the total loss. Their values are adjusted through experiments to balance the attention to shape and position.

[0041] The position loss function in this embodiment adopts the mean square error form, and a penalty coefficient is added to amplify the loss value of a larger position deviation. Assuming the true position is , and the predicted position is , the position loss function is: In the formula, i is the index of the sample point; N is the total number of sample points; is the position deviation penalty coefficient; is the position error threshold.

[0042] In this embodiment, the shape loss function uses the Dice similarity coefficient (Dice coefficient) to quantify the overlapping ratio of the actual shape region and the standard template, and a shape difference penalty term is added on the basis of the Dice coefficient. If the feature edges or key parts of the shape contour deviate by more than a specific threshold, additional penalties are introduced. In this embodiment, by adjusting the weight parameters and penalty coefficients, the proportion of position and shape in the total loss can be balanced, and the accuracy of the model for position and shape can be ensured. In this embodiment, assuming the predicted shape region is , and the true shape region , then the shape loss function is: In the formula, is the shape penalty amplification coefficient; is the degree of deviation at the edge or corner point; is the shape difference threshold, which is used to determine whether additional penalties are required for shape differences. When the feature edges or key parts of the shape contour deviate beyond this shape difference threshold, an additional penalty term is introduced.

[0043] During the model training process, this embodiment uses the preprocessed image data to train the model and continuously adjusts the model parameters using optimization algorithms to improve performance and accuracy. At the same time, techniques such as cross-validation and regularization are used to prevent overfitting and improve the generalization ability of the model. After training is completed, this embodiment uses the test data set to evaluate and test the model to verify its performance and accuracy. If the model performs poorly, it is possible to return and readjust steps such as data preprocessing, feature extraction, or algorithm selection. In the model evaluation, this embodiment sets multiple test scenarios to ensure that the model can accurately judge the accessibility of the interface unit under different angles, distances, and lighting conditions.

[0044] S5. When the matching verification passes, drive the target interface unit to establish an external physical connection with the chip bin and initiate data transmission.

[0045] S6. After the data transmission is completed, disconnect the external physical connection and control the brushless motor to drive the target interface unit to reset from the vertical working position to the horizontal storage position, preparing to execute the next interface switch.

[0046] Specifically, in this embodiment, the shape matching degree between the target interface unit and the standard template is calculated. The proportion of the shape overlapping area is quantified by the Dice similarity coefficient, and the shape matching between the target interface unit and the interface unit in the standard template is performed according to the proportion of the shape overlapping area. If the shape matching error is lower than the preset threshold, it is determined that the interface unit is in an accessible state and the shape matching passes. Otherwise, it is determined to be in an inaccessible state, and the position state of the target interface unit is readjusted. After the shape matching passes, the key corner point positions of the interface unit are detected in this embodiment. The Harris corner point algorithm is used to extract the corner coordinates, and the coordinate deviation from the standard template is calculated. If the position loss (including the over-threshold penalty term) meets the preset requirements, it is determined that the interface unit is in an accessible state and the position matching passes. Otherwise, it is determined to be in an inaccessible state, and the position state of the target interface unit is readjusted. In this embodiment, if the shape or position verification fails, the brushless motor is controlled to finely adjust the rotation angle of the interface unit until the matching error is within the allowable range. After the shape and position matching verifications pass, the interface unit is driven to establish a physical connection with the external device, and the data transmission protocol is started to complete the data interaction task. At the same time, during the data transmission process in this embodiment, the flight airflow is guided through the wedge-shaped gap of the metal shell and the chip bin of the interface unit to form a directional air film, reducing the contact surface thermal resistance. At the same time, the local temperature of the chip bin is detected in real time in this embodiment, and when it is detected that the local temperature of the chip bin exceeds the preset chip bin temperature threshold range, the brushless motor is actively controlled to drive the target interface unit to rotate, increasing the separation angle of the wedge-shaped gap between the target interface unit and the chip bin to strengthen the air film heat dissipation, and the unidirectional flow of the airflow generated during the flight of the drone is restricted through the linear airflow groove at the edge of the water droplet to improve the heat dissipation efficiency until the local temperature of the chip bin drops back within the preset chip bin temperature threshold range.

[0047] After the data transmission is completed, the external connection between the target interface unit and the chip bin is disconnected in this embodiment, and the brushless motor is controlled to drive the interface unit to reset from the vertical working position to the horizontal storage position, ending the current link task. During the reset process, the interface state is monitored in real time through the camera in this embodiment to ensure no mechanical interference. After the reset is completed, it is ready to execute the next interface switching process. Through electromechanical collaborative positioning, multi-scale visual verification, and dynamic heat dissipation control, this embodiment realizes the precise access and stable operation of the multi-interface link, solves the problems of low interface switching efficiency, poor anti-interference ability, and insufficient heat dissipation under high load of the traditional UAV autopilot, and ensures the reliability and real-time performance of the multi-task link in a complex environment.

[0048] For the specific limitations on the control method of a multi-interface multi-link UAV autopilot, reference can be made to the above limitations on a multi-interface multi-link UAV autopilot, which will not be elaborated here. Those of ordinary skill in the art can realize that the various modules and steps described in combination with the embodiments disclosed in the present application can be implemented in hardware, software, or a combination of both. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application. It should be noted that the magnitudes of the sequence numbers of the above processes do not mean the order of execution is prior or subsequent. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0049] The embodiment of the present invention provides a control method for a multi-interface multi-link UAV autopilot. The method includes: in response to a UAV docking request, determining a target interface unit according to the UAV docking request, and controlling the slider to axially move to the target interface unit for locking; driving the slider to drive the target interface unit to rotate from the horizontal storage position to the vertical working position, and collecting a target interface image of the target interface unit; preprocessing the target interface image to obtain a preprocessed interface image, and using a convolutional neural network to extract the target interface feature of the target interface unit from the preprocessed interface image; matching and verifying the target interface feature with the standard interface unit feature. When the matching verification fails, controlling the brushless motor to finely adjust the rotation angle of the target interface unit until the matching verification passes; when the matching verification passes, driving the target interface unit to establish an external physical connection with the chip bin and starting data transmission; after the data transmission is completed, disconnecting the external physical connection and controlling the brushless motor to drive the target interface unit to reset from the vertical working position to the horizontal storage position, preparing to execute the next interface switch. Compared with the prior art, this method realizes the precise access and stable operation of multi-interface links through electromechanical collaborative positioning, multi-scale visual verification, and dynamic heat dissipation control, significantly improving the multi-task adaptability and operation stability of the UAV autopilot.

[0050] The above embodiments only represent several preferred implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation to the scope of the invention patent. It should be pointed out that for those of ordinary skill in the art of this technology, without departing from the technical principle of the present invention, several improvements and substitutions can still be made, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the protection scope of the claims.

Claims

1. A multi-interface multi-link UAV autopilot, characterized by: It includes a multi-interface module and a chip compartment; wherein the multi-interface module includes a plurality of interface units, a brushless motor, a lead screw, a slider, a sliding sleeve and a stepping motor; The multi-interface module is used to control the dynamic access switching between the drone and the multi-link interface, and drives the lead screw to drive the slider to move axially through the stepping motor to select a target interface unit from multiple interface units, and controls the brushless motor to drive the sliding sleeve to rotate during the docking process, so that the target interface unit establishes a directional electrical connection with the CPU chip in the chip compartment through the electric slip ring contact structure, thereby suppressing electromagnetic crosstalk caused by parallel transmission of multiple interfaces; The chip bin is used to guide the airflow generated during the flight of the drone to form a directional air film through the wedge-shaped gap and the linear airflow groove on the edge of the water droplet, thereby reducing the contact thermal resistance between the chip bin and the interface unit.

2. The multi-interface multi-link UAV autopilot according to claim 1, characterized in that: Each of the interface units comprises an integrally formed metal shell and a rotating sleeve, and the sliding sleeve and the rotating sleeve are both provided with sliding grooves on their outer circumferences along the length direction; The rotating sleeve of each interface unit is rotatably connected to the outer periphery of the sliding sleeve through the sliding groove along the length direction of the sliding sleeve.

3. The multi-interface multi-link UAV autopilot according to claim 1, characterized in that: The output shaft of the brushless motor is fixedly connected to the sliding sleeve through the housing of the stepping motor, and the two ends of the sliding sleeve are rotatably connected to the top of the chip bin through a pair of sliding sleeve mounting seats, and the sliding sleeve rotates around the axis of the sliding sleeve mounting seat; The brushless motor is used to drive the stepping motor to drive the sliding sleeve to rotate, thereby driving the interface unit connected thereto to rotate to the working position.

4. The multi-interface multi-link UAV autopilot according to claim 2, characterized in that: The outer wall of the sliding sleeve and the inner wall of the rotating sleeve are both provided with an electric slip ring contact structure; The electric slip ring contact structure on the outer wall of the sliding sleeve is connected to the CPU chip in the chip bin through a data line, and when the selected target interface unit is rotated to a preset angle, the electric slip ring contact structure on the inner wall of the rotating sleeve contacts the electric slip ring contact structure on the outer wall of the sliding sleeve, so that the selected target interface unit forms a directional electrical connection with the CPU chip in the chip bin in a rotating state, thereby suppressing electromagnetic crosstalk when multiple interface units work in parallel.

5. The multi-interface multi-link UAV autopilot according to claim 4, characterized in that: The output shaft of the stepper motor is fixedly connected to the screw rod, and the screw rod is coaxially arranged in the sliding sleeve and threadedly connected to the slider; The stepper motor drives the screw rod to drive the slider to move axially in the sliding sleeve, and in the process of the slider sliding along the sliding sleeve, the slider cooperates with the sliding groove of the rotating sleeve of each interface unit in turn, so that the slider locks or unlocks different interface units in turn, thereby controlling the dynamic switching connection of the multi-link interface.

6. The multi-interface multi-link UAV autopilot according to claim 1, characterized in that: The chip bin comprises a rectangular box-shaped bin shell, in which a CPU chip and a battery are encapsulated; the upper surface of the bin shell is provided with a water droplet edge linear airflow groove running through the front and rear ends, and a wedge-shaped gap is formed between the front end of the upper surface of the bin shell and the same side of the interface unit; The water droplet edge linear airflow groove is used to guide the drone's flight airflow to flow unidirectionally in the front-to-back direction through the wedge-shaped gap to the chip bin and the interface unit when the drone is flying, and to generate a directional dynamic pressure air film through the dynamic pressure effect, thereby reducing the friction heat accumulation on the contact surface between the chip bin and the interface unit, thereby offsetting the temperature rise effect of the CPU chip under high load conditions.

7. The multi-interface multi-link UAV autopilot according to claim 1, characterized in that: Each interface unit is configured as at least four of a network interface, a USB interface, an I / O interface, an HDMI interface, a debugging interface and a power interface.

8. A control method for a multi-interface multi-link UAV autopilot, characterized in that: The control method comprises the following steps: In response to a drone docking request, a target interface unit is determined according to the drone docking request, and the slider is controlled to move axially to the target interface unit for locking; The driving slider drives the target interface unit to rotate from the horizontal storage position to the vertical working position, and collects the target interface image of the target interface unit; Preprocessing the target interface image to obtain a preprocessed interface image, and extracting target interface features of the target interface unit from the preprocessed interface image using a convolutional neural network; the convolutional neural network includes hybrid pooling, pyramid pooling, and residual network structures; Performing a matching check on the target interface characteristics and the standard interface unit characteristics, and when the matching check fails, controlling the brushless motor to fine-tune the rotation angle of the target interface unit until the matching check passes; When the matching check passes, the target interface unit is driven to establish an external physical connection with the chip compartment and start data transmission; After the data transmission is completed, the external physical connection is disconnected and the brushless motor is controlled to drive the target interface unit to reset from the vertical working position to the horizontal storage position, preparing to execute the next interface switching.

9. The control method of a multi-interface multi-link UAV autopilot as claimed in claim 8, characterized in that: The target interface features include target interface shape features and target interface position features; The step of matching and checking the target interface characteristics with the standard interface unit characteristics comprises: Perform shape matching on the target interface shape feature and the preset standard interface unit shape feature, and calculate the overlapping area ratio between the target interface shape feature and the preset standard interface unit shape feature based on the Dyce similarity coefficient to obtain the shape matching degree; Quantifying the shape matching error using a pre-constructed shape loss function; the shape loss function is composed of a Dyss similarity coefficient loss term and a shape deviation penalty term; If the shape matching error is lower than a preset shape error threshold range, it is determined that the matching check of the target interface shape feature has passed; Detecting the key corner point positions of the target interface unit according to the target interface position features, and extracting the corner coordinates from the key corner point positions; A position deviation penalty term and a Euclidean distance loss term are introduced to construct a position loss function, and the position loss function is used to quantify the position loss value corresponding to the coordinate deviation between the corner coordinate and the standard interface unit position feature; When the position loss value meets the preset position error threshold range, it is determined that the matching check of the target interface position feature has passed.

10. The control method of a multi-interface multi-link UAV autopilot as claimed in claim 8, characterized in that: The control method further comprises: During data transmission, the local temperature of the chip bin is detected in real time, and when it is detected that the local temperature of the chip bin exceeds the preset chip bin temperature threshold range, the brushless motor is actively controlled to drive the target interface unit to rotate, thereby increasing the wedge-shaped gap separation angle between the target interface unit and the chip bin, and constraining the airflow generated during the flight of the drone to flow in one direction through the linear airflow groove on the edge of the water droplet, until the local temperature of the chip bin drops back to the preset chip bin temperature threshold range.

Images