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

Abstract

The present invention relates to the field of drone technology, and more particularly to a multi-interface, multi-link drone autopilot and a control method thereof, comprising a multi-interface module for controlling the dynamic access switching between the drone and the multi-link interface, and a stepper motor driving a lead screw to drive the axial movement of a slider to select a target interface unit from a plurality of interface units. During the docking process, a brushless motor is controlled to drive a sliding sleeve to rotate, so that the target interface unit establishes a directional electrical connection with a CPU chip in a chip compartment through an electric slip ring contact structure, thereby suppressing electromagnetic crosstalk caused by parallel transmission of multiple interfaces. The chip compartment is used to guide the airflow generated by the drone during flight to form a directional air film through a wedge-shaped gap and a water droplet-edge linear airflow groove, thereby reducing the contact thermal resistance between the chip compartment and the interface unit. The present invention achieves efficient dynamic access switching between the drone and the multi-link interface and airflow-guided heat dissipation through the coordinated operation of the multi-interface module and the chip compartment, thereby improving the heat dissipation efficiency and operational stability of the system.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles (UAVs), and in particular to a multi-interface and multi-link UAV autopilot and a control method thereof. Background Art

[0002] With the continuous improvement of the integration level of civil UAV systems, the interaction requirements between their functional modules and ground systems are becoming increasingly complex. The requirements for multi-device collaborative communication and efficient heat dissipation have become key technical bottlenecks restricting system performance. In the existing UAV system architecture, the communication link between the flight control system and the ground station can no longer meet the needs of collaborative work of airborne equipment in multi-task scenarios. On the one hand, high-precision navigation equipment such as differential GPS mobile stations need to receive navigation message correction data uploaded by ground base stations in real time, while mission payload equipment needs to interact with ground control terminals for high-frequency status commands. The data interface types and protocol standards between different devices vary significantly, resulting in the need to configure multiple independent communication modules on the ground side, increasing the complexity of system integration and the difficulty of operation.

[0003] Traditional drone interface design has significant limitations. For example, in differential GPS systems, correction data is typically transmitted between ground base stations and airborne mobile stations using a point-to-point RF link. However, due to limitations in transmit power and antenna directivity, communication interruptions are prone to occur in complex terrain. Mission payload equipment often utilizes independent data links, whose interface specifications vary by manufacturer. This requires ground control terminals to be equipped with multiple physical interface conversion devices, increasing system weight and slowing down emergency response. When drones perform multi-task collaborative operations, the need for data synchronization between different devices necessitates the configuration of multiple independent receivers on the ground side, exacerbating the risk of electromagnetic interference and device power consumption. Furthermore, existing autopilot systems face severe thermal management challenges in terms of heat dissipation design. Current solutions often utilize a combination of passive cooling and natural convection, transferring heat from the CPU chip to the chassis via aluminum alloy heat sinks. However, with the increasing complexity of navigation algorithms and the demand for real-time processing, chip power density is increasing exponentially, making it difficult to control the temperature rise with traditional heat dissipation structures.

[0004] In summary, the limitations of existing interface and heat dissipation technologies have severely restricted the development of UAV systems. These technical bottlenecks not only reduce system reliability but also increase operation and maintenance costs, becoming a key obstacle restricting the expansion of UAVs into wider application areas. Therefore, the development of autopilot systems with both multi-interface intelligent management and highly integrated heat dissipation capabilities has become a core technical requirement for improving the overall performance of UAVs. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides a multi-interface and multi-link UAV autopilot and a control method thereof.

[0006] In a first aspect, the present invention provides a multi-interface multi-link UAV autopilot, comprising a multi-interface module and a chip compartment; wherein the multi-interface module comprises a plurality of interface units, a brushless motor, a lead screw, a slider, a sliding sleeve, and a stepping motor;

[0007] 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 stepper motor to select a target interface unit from multiple interface units. During the docking process, the brushless motor is controlled to drive the sliding sleeve to rotate, 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;

[0008] 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.

[0009] In a further embodiment, 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;

[0010] 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.

[0011] 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 seats;

[0012] 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.

[0013] In a further embodiment, 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;

[0014] 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 rotates 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.

[0015] In a further embodiment, the output shaft of the stepper motor is fixedly connected to the lead screw, and the lead screw is coaxially disposed in the sliding sleeve and threadedly connected to the slider;

[0016] 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.

[0017] In a further embodiment, the chip compartment includes a rectangular box-shaped compartment housing, wherein the CPU chip and the battery are encapsulated; the upper surface of the compartment housing 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 compartment housing and the same side of the interface unit;

[0018] 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 space between 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 frictional 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.

[0019] In a further implementation scheme, 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.

[0020] In a second aspect, the present invention provides a method for controlling a multi-interface and multi-link UAV autopilot, the method comprising the following steps:

[0021] In response to a drone docking request, determining a target interface unit according to the drone docking request, and controlling the slider to axially move to the target interface unit for locking;

[0022] 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;

[0023] 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;

[0024] Performing a matching check between the target interface characteristics and the standard interface unit characteristics, and if the matching check fails, controlling the brushless motor to fine-tune the rotation angle of the target interface unit until the matching check passes;

[0025] 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;

[0026] 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.

[0027] In a further embodiment, the target interface characteristics include target interface shape characteristics and target interface position characteristics; and the step of matching and verifying the target interface characteristics with the standard interface unit characteristics includes:

[0028] Performing shape matching on the target interface shape feature and the preset standard interface unit shape feature, and calculating 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;

[0029] Quantifying shape matching errors using a pre-constructed shape loss function; the shape loss function is composed of a Dice similarity coefficient loss term and a shape deviation penalty term;

[0030] 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 has passed;

[0031] Detecting key corner positions of a target interface unit according to the target interface position features, and extracting corner coordinates from the key corner positions;

[0032] Introducing a position deviation penalty term and a Euclidean distance loss term to construct a position loss function, and using 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;

[0033] When the position loss value satisfies a preset position error threshold range, it is determined that the matching verification of the target interface position feature has passed.

[0034] In a further embodiment, the control method further comprises:

[0035] During the data transmission process, the local temperature of the chip bin is detected in real time. 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 of the water droplet edge until the local temperature of the chip bin drops back to the preset chip bin temperature threshold range.

[0036] The present invention provides a multi-interface and multi-link unmanned aerial vehicle autopilot and a control method thereof. The unmanned aerial vehicle autopilot includes a multi-interface module and a chip compartment; 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 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 an electric slip ring contact structure, thereby suppressing electromagnetic crosstalk caused by parallel transmission of multiple interfaces; the chip compartment is used to guide the airflow generated during the flight of the unmanned aerial vehicle to form a directional air film through a wedge-shaped gap and a linear airflow groove with a water droplet edge, thereby reducing the contact thermal resistance between the chip compartment and the interface unit. Compared with the existing technology, the UAV autopilot realizes efficient dynamic access switching between the UAV and the multi-link interface through the coordinated work of the multi-interface module and the chip compartment. At the same time, the wedge-shaped gap and the linear airflow groove on the edge of the water droplet are combined to guide the airflow to form a directional air film, which reduces the contact thermal resistance between the chip compartment and the interface unit, improves the heat dissipation efficiency and operation stability of the system, and meets the high-performance communication and navigation control needs of the UAV in complex flight environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a schematic diagram of the structure of a multi-interface and multi-link UAV autopilot provided by an embodiment of the present invention;

[0038] Figure 2 1 is a schematic diagram of the cross-sectional structure of the UAV autopilot provided by an embodiment of the present invention;

[0039] Figure 3 This is a schematic diagram of the overall structure of the UAV autopilot provided by an embodiment of the present invention;

[0040] Figure 4 This is a schematic diagram of the chip bin structure provided by an embodiment of the present invention;

[0041] Figure 5 This is a flow chart of a method for controlling a multi-interface and multi-link UAV autopilot according to an embodiment of the present invention;

[0042] Figure 6 This is a flowchart of a method for controlling an unmanned aerial vehicle (UAV) autopilot according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following describes the embodiments of the present invention in detail with reference to the accompanying drawings. The embodiments are provided for illustrative purposes only and are not to be construed as limiting the present invention. The accompanying drawings are provided for reference and illustration only and do not constitute a limitation on the scope of protection of the present invention. Many changes may be made to the present invention without departing from the spirit and scope of the present invention.

[0044] refer to Figure 1 , the embodiment of the present invention provides a multi-interface multi-link UAV autopilot, such as Figure 1 As shown, the UAV autopilot includes a multi-interface module 1 and a chip compartment 2. The UAV autopilot is installed on a UAV 3 to realize the control and navigation of the UAV 3. In this embodiment, as shown in FIG. Figure 2 As shown, the multi-interface module 1 includes several interface units 101, a brushless motor 11, a screw 12, a slider 13, a sliding sleeve 14 and a stepper motor 15. The chip compartment serves as the computing and energy core of the autopilot. The chip compartment includes a rectangular box-shaped compartment shell, and the compartment shell is loaded with key components such as a CPU chip and a battery, ensuring the stable operation and long-term endurance of the drone autopilot in complex flight environments.

[0045] In some embodiments, 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 through the stepper motor to drive the slider to move axially 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 an electric slip ring contact structure, thereby suppressing electromagnetic crosstalk caused by parallel transmission of multiple interfaces.

[0046] like Figure 3 As shown, the multi-interface module 1 serves as the information interaction center of the drone autopilot. The multi-interface module integrates multiple interface units 101. Each interface unit 101 corresponds to a different data interface type to adapt to different data connection requirements. In some embodiments, 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 to meet diverse data transmission and control command reception requirements.

[0047] In this embodiment, each interface unit 101 includes an integrally formed metal shell 1011 and a rotating sleeve 1012. The sliding sleeve 14 serves as a key component for 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 fix and support the sliding sleeve 14 through the sliding sleeve mounting seats 21 to ensure that it is stable and does not shake. 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 sleeve 1012 of each interface unit 101 is rotatably connected to the outer periphery of the sliding sleeve through the sliding groove along the length direction of the sliding sleeve 14. Such a layout not only saves space, but also enables the interface unit 101 to flexibly adjust the position and angle as needed.

[0048] In order to further enhance the stability and reliability of the connection, the sliding sleeve 14 and the rotating sleeve 1012 are provided with sliding grooves 102 on the outer circumference along the length direction. 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 13 sliding along the sliding sleeve 14, the slider 13 cooperates with the sliding groove of the rotating sleeve of each interface unit in turn. As the slider 13 slides, the slider 13 locks or unlocks the sliding sleeve 14 and different interface units 101 in turn, thereby ensuring the stability and reliability of the interface unit 101 during the connection process, and at the same time realizes the control of dynamic switching connection of multi-link interfaces, thereby improving work efficiency.

[0049] In this embodiment, the brushless motor is used to drive the sliding sleeve to rotate by driving the stepper motor, thereby driving 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 stepper motor 15, and the housing of the stepper motor 15 is fixedly connected to the sliding sleeve 14, so that when the brushless motor 11 is started, the output shaft of the brushless motor 11 can drive the sliding sleeve 14 to rotate, thereby driving the interface unit 101 locked with the sliding sleeve 14 to rotate synchronously, thereby realizing flexible adjustment of the interface unit 101.

[0050] In this embodiment, a plurality of electric slip ring contact structures are provided on the outer wall of the sliding sleeve 14 and the inner wall of the rotating sleeve 1012. The electric slip ring contact structure on the outer wall of the sliding sleeve 14 is connected to the CPU chip in the chip compartment 2 through a data line, ensuring stable signal transmission. In this embodiment, when the metal shell 1011 of the interface unit 101 is attached to the compartment shell 201 of the chip compartment 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, the interface unit 101 and the CPU chip are in a state of contact. No electrical connection is formed between the two interfaces. However, when the selected target interface unit 101 rotates to a preset angle, the electrical slip ring contact structure on the outer wall of the sliding sleeve 14 contacts the electrical slip ring contact structure on the inner wall of the rotating sleeve 1012, thereby achieving a directional electrical connection between the selected target interface unit and the CPU chip in the rotating state, suppressing electromagnetic crosstalk when multiple interface units operate in parallel. It should be noted that the electrical slip ring structure provided at the contact points between the outer wall of the sliding sleeve 14 and the inner wall of the rotating sleeve 1012 is similar to an electrical slip ring. An electrical slip ring allows a mechanical structure to maintain a stable electrical connection while rotating 360 degrees. Therefore, through this design, this embodiment not only enables flexible rotation of the interface unit 101, but also ensures reliable data transmission and power supply between the interface unit 101 and the CPU chip during rotation, providing a strong guarantee for the normal operation of the device.

[0051] In this embodiment, a wedge-shaped gap 202 is provided between the front end of the upper surface of the hopper shell 201 and the front end of the lower surface of the metal shell 1011, thereby utilizing the airflow generated by the drone 3 during flight to enhance the heat dissipation effect of the CPU chip. The water droplet edge linear airflow groove is used to guide the drone's flight airflow through the wedge-shaped gap to flow unidirectionally in the front-to-back direction to the space between the chip hopper and the interface unit during flight. A directional dynamic pressure air film is generated through the dynamic pressure effect, thereby reducing the frictional heat accumulation on the contact surface between the chip hopper and the interface unit, thereby offsetting the temperature rise effect of the CPU chip under high load conditions. Figure 4 This is a schematic diagram of the chip warehouse structure.

[0052] Specifically, when the drone 3 is flying, airflow is generated, and the airflow enters between the chip bin 2 and the interface unit 101 through the wedge-shaped gap 202. Based on the fluid dynamic 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 isolating role, but also reduces the thermal resistance of the contact surface and enhances the heat dissipation efficiency of the CPU chip. In order to further optimize the flow path of the airflow, this embodiment is provided with an airflow groove 203 on the upper surface of the bin shell 201. The airflow groove 203 is set through the front and back directions of the bin shell 201. The airflow groove 203 is composed of a number of water droplet edge line shaped airflow groove units connected to each other, and the shape of each airflow groove unit is a water droplet edge structure with a large front and a small back. The edge line shape and the setting of the airflow groove 203 not only ensure that the air film is stably opened, but also can guide the airflow to flow back and forth along a specific direction. The airflow will flow in from one side of the wedge-shaped gap 202 set between the chip compartment 2 and the interface unit 101, and then flow to the other side along the direction of the airflow groove 203, that is, the airflow is guided to flow unidirectionally from the side of the wedge-shaped gap 202 to the other side. This directional flow of airflow not only improves the heat dissipation efficiency, but also ensures the stability and controllability of the entire heat dissipation process. Therefore, this embodiment realizes an efficient and stable CPU chip heat dissipation solution by arranging the wedge-shaped gap 202 and the airflow groove 203, which not only improves the overall performance of the drone, but also provides a strong guarantee for its long-term stable operation.

[0053] In one embodiment, Figure 5 As shown, an embodiment of the present invention provides a method for controlling a multi-interface and multi-link UAV autopilot, the control method comprising the following steps:

[0054] S1. In response to a drone docking request, determine a target interface unit according to the drone docking request, and control the slider to move axially to the target interface unit for locking.

[0055] S2. 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.

[0056] Specifically, such as Figure 6 As shown, this embodiment responds to the drone docking request, determines the interface type required for docking of the drone according to the communication protocol between the drone and the drone autopilot, and determines the target interface unit according to the interface type required for docking of the drone, and then controls the stepper motor to drive the slider to move to the sliding slot position of the corresponding interface. The rotation of the brushless motor drives the slider to further rotate, so that the selected target interface unit is transformed from a horizontal state to a 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 its own camera to capture the target interface image of the target interface unit to achieve accurate access to the interface unit.

[0057] S3. Preprocess the target interface image to obtain a preprocessed interface image, and use a convolutional neural network to extract the target interface features of the target interface unit from the preprocessed interface image; the convolutional neural network includes hybrid pooling, pyramid pooling and residual network structures.

[0058] This embodiment transfers the collected target interface image to the CPU chip, and performs preprocessing operations on the target interface image to obtain a preprocessed interface image, wherein the preprocessing operations include image segmentation, filtering, enhancement and denoising operations to improve the quality and stability of the image. In this process, this embodiment adopts preprocessing tools such as median filtering, Gaussian filtering, morphological operations, binarization or histogram equalization to effectively remove noise in the image, enhance target features, and ensure that the image quality meets 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 perform feature extraction on the preprocessed image according to the task type and data characteristics. At the same time, in order to achieve more refined feature extraction, this embodiment adopts an improved convolutional neural network structure. In the target interface unit shape recognition process, this embodiment adopts a multi-scale convolution kernel structure. At the same The convolutional layer uses three different sizes of convolution kernels: 3x3 small size, 5x5 medium size and 7x7 large size to capture the different scale features of the interface units. Among them, the 3x3 small size convolution kernel can accurately capture the detailed features of the target interface unit, such as edges, corners, etc., and enhance the sensitivity to small position changes; the 5x5 medium size convolution kernel is mainly used to capture the medium-scale features of the interface unit, such as large contour changes and local shapes, and improve the ability to extract contour and angle information; the 7x7 large size convolution kernel can capture the global features of the interface unit, reduce the impact of perspective transformation and scaling on shape recognition, and enhance the model's perception of the overall structure. Large size convolution kernels are usually placed in the shallower layers of the network or the downsampled image layer to extract a wide range of features and capture the overall structural characteristics of the interface, so that they can simultaneously capture the local and global features of the interface unit, provide stable recognition effects for interface shapes in different environments, and make correct judgments and processing.

[0059] The convolutional neural network structure in this embodiment achieves more refined feature extraction through multi-layer convolution and pooling operations. After the convolution layer, this embodiment introduces a hybrid pooling strategy in the pooling process to achieve better adaptation to features of different scales and details. The hybrid pooling strategy alternately uses maximum pooling and average pooling in different convolution layers. Maximum 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, this embodiment uses pyramid pooling in higher-level feature maps, divides the higher-level feature maps into regions of different scales, and pools each region separately to capture contextual information at different scales, so that the model can perceive shape features at different scales, thereby enhancing the model's ability to recognize the overall and local features of interface units.

[0060] In addition, this embodiment introduces a residual network structure to ensure that more details and position information are retained during the feature extraction process. Each residual block not only contains convolution + batch normalization + ReLU activation operations, but also adds a skip connection (SkipConnection). Through the skip connection, more original details and position information are retained during the feature propagation process. This embodiment adopts a bottleneck structure of 1x1, 3x3, and 1x1 in the residual block, among which 1x1 convolution is used for dimensionality reduction and dimensionality increase to reduce the amount of calculation; 3x3 convolution is used for feature extraction, which retains important feature details while reducing the amount of calculation. The skip connection retains the original information during the feature propagation process to avoid detail loss, further improving the recognition accuracy and reducing the model complexity.

[0061] In terms of position recognition, this embodiment introduces a hierarchical classification model based on the convolutional neural network structure. The hierarchical classification model judges the accessibility of the interface unit in steps. The first-layer model is used to judge whether the shape matching of the interface unit meets the access requirements, and the second-layer model verifies the accuracy of the interface unit position. This hierarchical classification model distribution verification setting not only reduces the misjudgment rate, but also significantly improves the real-time performance of recognition, thereby realizing accurate recognition of the shape and position of the interface unit, providing a strong guarantee for the precise docking of equipment such as drone autopilots.

[0062] S4. Perform matching verification on the target interface characteristics and the standard interface unit characteristics. If the matching verification fails, control the brushless motor to fine-tune the rotation angle of the target interface unit until the matching verification passes.

[0063] In some embodiments, the target interface characteristics include target interface shape characteristics and target interface position characteristics; and the step of matching and verifying the target interface characteristics with standard interface unit characteristics includes:

[0064] Performing shape matching on the target interface shape feature and the preset standard interface unit shape feature, and calculating 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;

[0065] Quantifying shape matching errors using a pre-constructed shape loss function; the shape loss function is composed of a Dice similarity coefficient loss term and a shape deviation penalty term;

[0066] 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 has passed;

[0067] Detecting key corner positions of a target interface unit according to the target interface position features, and extracting corner coordinates from the key corner positions;

[0068] Introducing a position deviation penalty term and a Euclidean distance loss term to construct a position loss function, and using 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;

[0069] When the position loss value satisfies a preset position error threshold range, it is determined that the matching verification of the target interface position feature has passed.

[0070] Specifically, this embodiment extracts the target interface features of the target interface unit 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 to extract 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 to evaluate the degree of matching. If the matching error is within a preset threshold range, the edge check is determined to be passed; if the error is too large, the target interface unit is determined to be inaccessible. On this basis, this embodiment further detects subtle features of key parts on the edge, such as specific bending points, concave and convex parts, and other key parts, and controls the errors of these key parts within a stricter threshold to improve the accuracy of recognition.

[0071] At the same time, this embodiment uses Harris corner detection or Shi-Tomasi corner detection algorithm to identify the key corner point 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 grayscale change, compares the detected corner point positions with the corner point positions in the standard template, and calculates the corner point deviation. If the matching of the edge and the corner point meets the preset requirements, it is determined that the subtle feature verification of the target interface unit has passed. At this time, the target interface unit is accurate and accessible. Otherwise, this embodiment will determine that the target interface unit is inaccessible and readjust the position of the drone for re-verification.

[0072] Based on feature extraction and verification, this embodiment compares the extracted features with known data and classifies the new data into corresponding categories according to pre-set classification rules. Then, based on system requirements, the data is analyzed and inferred using experience, rules, or statistical methods. In order to improve the matching accuracy of shape and position, this embodiment designs a dedicated comprehensive loss function. The comprehensive loss function consists of two parts: a position loss function and a shape matching loss function. Penalties for position deviation and shape matching are added to make the model pay more attention to the accuracy of shape and position during training. The comprehensive loss function is specifically expressed as:

[0073]

[0074] Where, is the comprehensive loss function; is the position loss function; is the shape matching loss function; and Both are weight parameters, which are used to control the proportion of position and shape loss in the total loss. Their values ​​are adjusted through experiments to balance the attention of shape and position.

[0075] The position loss function in this embodiment adopts the mean square error form and adds a penalty coefficient to amplify the loss value of large position deviation. Assuming that the true position is , the predicted position is , the position loss function is:

[0076]

[0077] Where 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.

[0078] In this embodiment, the shape loss function uses the Dice similarity coefficient (Dice coefficient) to quantify the overlap ratio between the actual shape area and the standard template, and adds a shape difference penalty term on the basis of the Dice similarity coefficient. If the characteristic edge or key part of the shape contour deviates by more than a specific threshold, an additional penalty is introduced. By adjusting the weight parameters and the penalty coefficient, this embodiment can balance the proportion of position and shape in the total loss and ensure the accuracy of the model for position and shape. This embodiment assumes that the predicted shape area is , the true shape area , then the shape loss function is:

[0079]

[0080] Where, is the penalty coefficient for amplified shape; is the degree of deviation at the edge or corner; is a shape difference threshold, which is used to determine whether the shape difference requires additional penalty. When the characteristic edge or key part of the shape contour deviates beyond the shape difference threshold, an additional penalty term is introduced.

[0081] During the model training process, this embodiment uses preprocessed image data to train the model, and uses optimization algorithms to continuously adjust model parameters to improve performance and accuracy. At the same time, cross-validation, regularization and other techniques are used to prevent overfitting and improve the generalization ability of the model. After the training is completed, this embodiment uses a test data set to evaluate and test the model to verify its performance and accuracy. If the model performs poorly, it can return to readjust the data preprocessing, feature extraction or algorithm selection steps. 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.

[0082] S5. 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.

[0083] S6. 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.

[0084] Specifically, this embodiment calculates the shape matching degree between the target interface unit and the standard template, quantifies the shape overlapping area ratio by the Dyss similarity coefficient, and performs shape matching between the target interface unit and the interface unit in the standard template according to the shape overlapping area ratio. If the shape matching error is lower than the preset threshold, the interface unit is determined to be in an accessible state and the shape matching passes. Otherwise, it is determined to be inaccessible and the position state of the target interface unit is readjusted. After the shape matching passes, this embodiment detects the position of the key corner points of the interface unit, extracts the corner coordinates by using the Harris corner algorithm, and calculates the coordinate deviation from the standard template. If the position loss (including the over-threshold penalty item) meets the preset requirements, the interface unit is determined to be in an accessible state and the position matching passes. Otherwise, it is determined to be inaccessible and the position state of the target interface unit is readjusted. In this embodiment, if the shape or position check fails, Then the brushless motor is controlled to fine-tune the rotation angle of the interface unit until the matching error is within the allowable range. After the shape and position matching verification is passed, the interface unit is driven to establish a physical connection with the external device, the data transmission protocol is started, and the data interaction task is completed. At the same time, during the data transmission process, the present embodiment guides the flying airflow to form a directional air film through the wedge-shaped gap between the metal shell of the interface unit and the chip bin, thereby reducing the thermal resistance of the contact surface. At the same time, the present embodiment detects the local temperature of the chip bin 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, and the wedge-shaped gap separation angle between the target interface unit and the chip bin is increased to enhance the heat dissipation of the air film. The linear airflow grooves at the edges of the water droplets constrain the airflow generated during the flight of the drone to flow in one direction, thereby improving the heat dissipation efficiency, until the local temperature of the chip bin drops back to within the preset chip bin temperature threshold range.

[0085] After the data transmission is completed, this embodiment disconnects the external connection between the target interface unit and the chip compartment and controls the brushless motor 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, this embodiment monitors the interface status in real time through the camera to ensure that there is no mechanical interference. After the reset is completed, it is ready to execute the next interface switching process. This embodiment realizes the precise access and stable operation of multi-interface links through electromechanical collaborative positioning, multi-scale visual verification and dynamic heat dissipation control, solving the problems of low interface switching efficiency, poor anti-interference and insufficient high-load heat dissipation of traditional UAV autopilots, ensuring the reliability and real-time performance of multi-task links in complex environments.

[0086] For the specific definition of a control method for a multi-interface multi-link UAV autopilot, please refer to the above-mentioned definition of a multi-interface multi-link UAV autopilot, which will not be repeated here. A person of ordinary skill in the art will appreciate that the various modules and steps described in conjunction with the embodiments disclosed in this application can be implemented in hardware, software, or a combination of both. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application. It should be noted that the size of the sequence number of each of the above processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0087] An embodiment of the present invention provides a control method for a multi-interface and multi-link unmanned aerial vehicle autopilot, the method comprising: responding to a drone docking request, determining a target interface unit according to the drone docking request, and controlling a slider to move axially to the target interface unit for locking; driving the slider to rotate the target interface unit from a horizontal storage position to a 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 extracting target interface features of the target interface unit from the preprocessed interface image using a convolutional neural network; matching and verifying the target interface features with standard interface unit features, and when the matching verification fails, controlling a brushless motor to fine-tune 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 a chip compartment 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 switching. Compared with the existing technology, this method achieves 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 operational stability of the UAV autopilot.

[0088] The above-described embodiments merely represent several preferred implementations of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art could make several improvements and substitutions without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be based on the scope of protection 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 several interface units, a brushless motor, a lead screw, a slider, a sliding sleeve and a stepping motor; each of the interface units includes an integrally formed metal shell and a rotating sleeve; 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 stepper motor to select a target interface unit from multiple interface units. During the docking process, the brushless motor is controlled to drive the sliding sleeve to rotate, so that the target interface unit establishes a directional electrical connection with the CPU chip in the chip compartment through an electric slip ring contact structure, thereby suppressing electromagnetic crosstalk caused by parallel transmission of multiple interfaces. The outer wall of the sliding sleeve and the inner wall of the rotating sleeve are both provided with electric slip ring contact structures. The electric slip ring contact structure on the outer wall of the sliding sleeve is connected to the CPU chip in the chip compartment via a data line. When the selected target interface unit rotates 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 compartment in a rotating state, thereby suppressing electromagnetic crosstalk when multiple interface units operate in parallel. The chip bin is used to guide the airflow generated during the flight of the drone to form a directional air film through a wedge-shaped gap and a water droplet-edge linear airflow groove, thereby reducing the contact thermal resistance between the chip bin and the interface unit. The chip bin includes 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 space between 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 frictional 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.

2. The multi-interface, multi-link UAV autopilot according to claim 1, characterized in that: 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 stepper motor. 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 seats. 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 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.

5. 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.

6. A control method for a multi-interface multi-link UAV autopilot based on any one of claims 1-5, characterized in that: The control method comprises the following steps: In response to a drone docking request, determining a target interface unit according to the drone docking request, and controlling the slider 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 between the target interface characteristics and the standard interface unit characteristics, and if 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.

7. The method for controlling a multi-interface and multi-link UAV autopilot according to claim 6, wherein: The target interface features include target interface shape features and target interface position features; The step of matching and verifying the target interface characteristics with the standard interface unit characteristics comprises: Performing shape matching on the target interface shape feature and the preset standard interface unit shape feature, and calculating 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 shape matching errors using a pre-constructed shape loss function; the shape loss function is composed 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 has passed; Detecting key corner positions of a target interface unit according to the target interface position features, and extracting corner coordinates from the key corner positions; Introducing a position deviation penalty term and a Euclidean distance loss term to construct a position loss function, and using 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 satisfies a preset position error threshold range, it is determined that the matching verification of the target interface position feature has passed.

8. The method for controlling a multi-interface and multi-link UAV autopilot according to claim 6, wherein: The control method further includes: During the data transmission process, the local temperature of the chip bin is detected in real time. 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 of the water droplet edge until the local temperature of the chip bin drops back to the preset chip bin temperature threshold range.

Images