A method and system for unmanned tractor to realize autonomous indoor and outdoor detachment

Abstract

The present invention discloses a method and system for an unmanned tractor to realize autonomous detachment indoors and outdoors. Based on a factor graph optimization fusion framework, the position and posture of the unmanned vehicle is used as an optimization variable, and is divided into unary edges, binary edges, and multiary edges according to the number of optimization variables bound by constraints. By fusing observation information from multiple sensors, high-precision autonomous positioning of the unmanned tractor is realized. Based on the above positioning information, a drivable global path for the unmanned tractor is planned on the road from the apron to the air cargo terminal. At the same time, in the hook area, according to the current coordinates of the vehicle, the high emissivity characteristics of the reflective sticker are used, combined with its relative geometric position on the tractor, and the position of the towing ring of the towing vehicle is calculated through coordinate conversion, and a hook reversing path is planned. The present invention realizes the accuracy and operation efficiency of the hooking, reduces manual intervention, and makes the application scenarios of unmanned tractors more common and wider.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned driving, and in particular relates to a method and a system for an unmanned driving tractor to realize autonomous detachment indoors and outdoors. Background Art

[0002] Autonomous navigation and positioning technology is the core of unmanned driving technology and the basic guarantee for its application in various industries. Navigation and positioning technology in outdoor environments mainly relies on the interaction between satellites, ground base stations and receivers to achieve accurate positioning of targets. At present, the global positioning system GPS or Beidou satellite navigation system BDS has developed quite maturely, and the positioning accuracy can reach the centimeter level in open scenes. However, in weak satellite scenes, such as indoor and outdoor transition areas in airport environments, between buildings, airport bridges, underground baggage carousels, etc., it is impossible to obtain continuous and reliable positioning results for a certain positioning source.

[0003] A pallet truck is used to load cargo. It has a towing rod and a towing ring at the front end and a hooking device at the rear end. Currently, the hooking operation is mainly done manually. When it comes to autonomous driving, there is no operator to manually hook the autonomous driving tractor and the pallet truck. If a separate hooking operator is set up, the labor cost will increase and the meaning of autonomous driving will be lost. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method and system for an unmanned tractor to achieve autonomous detachment indoors and outdoors. By fusing observation information from multiple sensors, high-precision autonomous positioning of the unmanned tractor can be achieved even in an environment with poor or no satellite signals.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for an unmanned tractor to realize autonomous detachment indoors and outdoors, the method comprising:

[0007] Step 1: Build a fusion framework autonomous positioning system based on factor graph optimization, calculate the real-time posture information of the unmanned tractor based on satellite position constraints, IMU pre-integration constraints, and lidar constraints to obtain real-time positioning information;

[0008] Step 2: equip the unmanned tractor with an autonomous detaching device and equip the trailer with a traction ring, wherein the autonomous detaching device can identify the traction ring through a laser radar;

[0009] Step 3: Based on the real-time positioning information of the unmanned tractor, plan the global path of the unmanned tractor and the local reversing path of the hooking area;

[0010] Step 4: In combination with the real-time posture information and the planned local reversing path of the hooking area, the autonomous unhooking device performs autonomous hooking and unhooking operations with the traction ring.

[0011] In another aspect, the present invention provides a device for realizing autonomous detachment of an unmanned tractor indoors or outdoors, comprising:

[0012] The real-time positioning information unit is used to build a fusion framework autonomous positioning system based on factor graph optimization. Based on satellite position constraints, IMU pre-integration constraints, and lidar constraints, the real-time posture information of the unmanned tractor is calculated to obtain real-time positioning information;

[0013] An assembly unit, used to assemble an autonomous detaching device for the unmanned tractor and a traction ring for the trailer, wherein the autonomous detaching device can identify the traction ring through a laser radar;

[0014] A path planning unit, used to plan the global path of the unmanned tractor and the local reversing path of the hooking area based on the real-time positioning information of the unmanned tractor;

[0015] The autonomous operation unit is used to combine the real-time posture information and the planned local reversing path of the hooking area. The autonomous unhooking device performs autonomous hooking and unhooking operations with the traction ring.

[0016] In a third aspect, the present invention provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned method for an unmanned tractor to achieve autonomous detachment indoors and outdoors.

[0017] In a fourth aspect, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enables the processor to implement the aforementioned method for an unmanned tractor to achieve autonomous detachment indoors and outdoors.

[0018] The beneficial effects of the present invention are:

[0019] The present invention can greatly improve the accuracy of unmanned tractor positioning in indoor and outdoor environments, especially when the satellite signal is poor or no signal. At the same time, by utilizing the high reflection intensity of the reflective sticker under the laser radar, the trailer towing ring is integrated with detection and positioning, thereby achieving the accuracy and operation efficiency of the hooking, reducing manual intervention, making the operation process of unmanned tractor in indoor / outdoor environments smoother, and making the application scenarios of unmanned tractor more common and wider. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1This is the principle diagram of the indoor and outdoor autonomous positioning framework of the unmanned tractor;

[0021] Figure 2 This is the principle diagram of the self-positioning algorithm of the fusion framework based on factor graph optimization;

[0022] Figure 3 This is the principle diagram of the two-stage lidar mileage calculation method based on the optical flow algorithm;

[0023] Figure 4 Schematic diagram of path planning for unmanned tractor in airport environment;

[0024] Figure 5 This is a schematic diagram of the automatic unhooking device and sensor installation for the unmanned tractor;

[0025] Figure 6 This is the electrical block diagram of the automatic unhooking hardware system;

[0026] Figure 7 This is the mechanical structure diagram of the automatic unhooking device;

[0027] Figure 8 Design a flow chart for the autonomous hook perception module software;

[0028] Fig. 9 The original point cloud image and calculation result image of the pallet truck detected by radar;

[0029] Fig.10 Automatic hooking flow chart for unmanned tractor;

[0030] Fig.11 Flowchart for automatic uncoupling of unmanned tractor. DETAILED DESCRIPTION

[0031] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0032] like Figure 1 As shown, the unmanned tractor relies on the global positioning system GPS to achieve its navigation and positioning in an open outdoor environment, and its positioning accuracy can reach the centimeter level. However, in weak satellite scenarios, such as indoor and outdoor transition areas in airport environments, between buildings, airport bridges, underground baggage carousels, etc., it is impossible to obtain continuous and reliable positioning results by relying solely on a certain positioning source. In the fusion framework based on factor graph optimization designed by the present invention, by fusing observation information from multiple sensors, high-precision autonomous positioning of the unmanned tractor can be achieved in an environment with poor satellite signals or no signals.

[0033] like Figure 2As shown in the figure, the fusion framework autonomous positioning system based on factor graph optimization takes the position and posture of the unmanned vehicle as the optimization variable, and divides it into one-dimensional edges, two-dimensional edges and multi-dimensional edges according to the number of optimization variables bound by constraints, such as satellite position constraints as one-dimensional edges, IMU pre-integration, laser radar odometer, loop closure and other constraints as two-dimensional edges. By fusing observation information from multiple sensors, high-precision autonomous positioning of the unmanned vehicle can be achieved in an environment with poor or no satellite signal.

[0034] The observation equation of the binary edge constraint is: ,in 6 degrees of freedom , =1,2, where is the roll angle, is the pitch angle, is the heading angle, , , In the inertial navigation coordinate system , , The coordinate values ​​of , from the perspective of least squares, can be written as: If other observations are added, the overall cost function can be obtained as follows:

[0035] ,

[0036] in, , Indicates the number of optimized poses.

[0037] The optimization equation is solved by the Gauss-Newton method. When the unmanned vehicle's position is updated, the RPY (Roll-Pitch-Yaw) angle needs to be maintained. After multiple iterations, the optimized pose can be obtained.

[0038] The IMU pre-integration constraint is widely used in the field of positioning. Due to its high frequency and short-term reliability, integrating it into the mapping system will bring higher positioning accuracy and better map details. Since the measurement frequency of IMU is too high, in order to prevent the optimization process from wasting too many resources of the unmanned vehicle computing platform with high-frequency data calls, the pre-integration method is used to accumulate the data for unified processing during the implementation process.

[0039] The satellite position constraint is to build a self-positioning system for unmanned vehicles that is available in the entire domain. When designing the self-positioning module, the satellite is also added as a constraint to the posture fusion module (a submodule of the self-positioning module). In the process of using satellite information, the status of satellite information is judged in real time. Only when its data status is normal will it be added to the posture fusion module. Satellite constraints can correct the self-positioning of the system when the satellite signal is normal, which can greatly reduce the accumulated positioning error of the system and effectively improve the positioning accuracy of the system.

[0040] Since the multi-sensor fusion framework integrates data from many sensors for optimization and solution, the output of the optimized posture is difficult to achieve real-time, but the positioning output of the IMU is reliable in the short term. Based on the output posture of the current multi-sensor fusion framework, the sensor data of the IMU can be integrated to output high-frequency positioning information.

[0041] In the specific implementation of the posture fusion module, two threads can be opened: Thread 1 receives the optimized posture output by the multi-sensor fusion framework; Thread 2 receives the output of Thread 1 and the data of the IMU, and fuses the two for high-frequency output. Through multi-threading, a real-time and efficient self-positioning system for unmanned tractors can be built.

[0042] like Figure 3 As shown, the laser radar odometer constraint can add point cloud registration constraints to the point cloud data generated by the laser radar. In order to quickly and efficiently obtain the point cloud registration result to adapt to the computing resource limitations of the unmanned vehicle, the present invention designs a two-stage laser radar odometer based on the optical flow algorithm, and the steps are:

[0043] S101: Extract features from the point cloud generated by the laser radar. The curvature calculation method of the point cloud can be defined as:

[0044] ,

[0045] in, is the laser radar coordinate system ( Department) Line Column of laser points, for The surrounding laser point collection;

[0046] S102: In order to ensure that the correct data will dominate the optimization and reduce the impact of misjudgment, the curvature calculated in step S101 is weighted according to the distance and filtered to obtain the effective value of the curvature;

[0047] S103: judging whether the laser point is a line feature or a plane feature according to the size of the effective value of the optimized curvature; if the curvature is large and exceeds the set threshold, it can be regarded as a line feature, otherwise it can be regarded as a surface feature;

[0048] S104: Using the feature points obtained in the previous step, project them onto a two-dimensional bird's-eye view, and use 2D images instead of 3D points to store the point cloud. Direction information is lost, but it can be found in 2D registration in a plane can balance registration accuracy and computation time;

[0049] S105: Exploitation The optical flow method performs fast coarse registration of two-dimensional sparse point clouds, obtains matching pairs of points between two consecutive frames, and effectively improves the computational efficiency of point cloud registration;

[0050] S106: After obtaining the matching feature point pairs in the previous step, the objective function of the coarse registration can be defined as:

[0051] ,

[0052] in, is the number of point clouds projected into the bird's-eye view, is the rotation matrix, is the translation matrix, and is the corresponding feature point between two frames obtained by the optical flow method, is the coarse registration transformation matrix;

[0053] S107: Use optical flow to obtain the rough registration result, and guide the fine registration of the three-dimensional sparse point cloud based on NDT. The rough registration result of the plane is obtained by The registration rotation angle and registration translation angle of the plane are set to zero. After several iterations, the final fine registration result can be obtained, and then the final LiDAR posture constraint result can be obtained.

[0054] like Figure 4 As shown, on the basis of being able to realize autonomous positioning of the unmanned tractor, high-precision positioning information is used to perform path planning for the unmanned tractor, including global path planning and local reversing path planning in the hooking area.

[0055] The global path planning of the unmanned tractor is to plan a driving path for the unmanned tractor on the drivable road of the tractor from the apron to the baggage carousel of the air cargo terminal at the airport. The specific waypoint format is shown in Table 1 below.

[0056] Table 1

[0057]

[0058] The latitude, longitude and heading information are provided by the indoor and outdoor positioning information obtained by the aforementioned autonomous positioning system.

[0059] The road section types include on-road, pre-intersection, intersection, hook-up area and unhooking area. The speed limits are different in different sections. For example, the upper speed limit on the road is 20km / h, and the upper speed limit in the unhooking / unhooking area is 10km / h.

[0060] The types of lane changes are different on different road sections. For example, changing lanes and maneuvering are only allowed when encountering obstacles on the road. Otherwise, the vehicle must stop when encountering obstacles.

[0061] The types of parking points are divided into non-stop points, road stop points, hook stop points, and unhooked stop points. Different types of parking have different actions after parking. After parking at a hook stop point, the vehicle will be automatically hooked. After parking at an unhooked stop point, the vehicle will be automatically unhooked. After parking at a road stop point, the vehicle will only stop. The autonomous driving vehicle itself will not take any action. It will only stop and wait, and then continue driving.

[0062] The initial status of the stop point is 0, that is, not executed. When driving to a stop point, the flag position is set to 1, that is, it has been executed. After reversing in the hook area and passing the hook stop point again, the parking task will not be repeated.

[0063] The uncoupling / uncoupling area includes a tractor parking area, a luggage trolley parking area and an uncoupling / uncoupling parking point. The tractor parking area and the luggage trolley parking area are kept in a straight line as much as possible during the planning process to reduce lateral errors during the uncoupling process.

[0064] The local reversing path planning of the hooking area of ​​the unmanned tractor is that after the unmanned tractor stops at the stop point in the hooking area, the perception system detects and identifies the traction ring of the trailer and sends its coordinate information to the decision system. The decision system plans a reversing path by two-point linear interpolation method according to the positioning coordinates of the traction ring and the current coordinates of the vehicle, and slightly extends it to ensure that the traction ring of the trailer can accurately enter the guide groove of the tractor and trigger the limit switch signal after sufficient collision.

[0065] like Figure 5 As shown, the present invention uses a laser radar plus a high-strength reflective strip to identify the trailer. In the embodiment, an automatic driving tractor capable of providing power is used, a decoupling device is provided at the bottom of the rear end, and a laser radar is provided above the decoupling device at the middle end of the rear, and the reflective strip is identified by the laser radar.

[0066] The end of the towing rod in the middle of the front end of the trailer that docks with the unhooking device of the tractor is provided with a trailer hook point corresponding to the tractor, namely the towing ring. In the hook parking area, a hook stop point is set to ensure the accuracy of the hook. When placing the trailer, the towing ring is placed on the hook stop point.

[0067] The front end of the pallet truck is provided with a left reflective strip and a right reflective strip. In the embodiment, with the towing ring of the pallet truck as the center, reflective strips are symmetrically attached at 1m on both sides of the towing rod. The size of the reflective strip is 5cm. 25cm, to ensure that the laser radar can reach the reflective strip, thereby realizing the recognition of the reflective strip.

[0068] Among them, the identification of reflective strips is achieved through the combination of luminous rate and installation mode. The understanding of the reflective rate is that objects of different materials have different reflective rates to the laser radar. According to this principle, this example uses laser reflective strips made of materials with specific reflective rates, and filters the data to find the position according to the specific reflective rate in the radar data. The installation mode of the reflective strip is based on the length of the towing ring, the towing rod and the geometric position relationship of the reflective strip. The position of the preset reflective strip can be filtered out from these known information, the interference point can be removed, and then the hook parking point, that is, the relative coordinate position of the towing ring relative to the towing vehicle, can be calculated.

[0069] like Figure 6 As shown, the integrated controller mainly receives the autopilot command issued by the upper computer autopilot system through the CAN network, such as throttle, steering, braking, gear, etc., and forwards it to the lower-level actuator; at the same time, by detecting the limit sensor signal, it controls the unhooking motor to operate the pull rod upward and downward, guides the traction pin to enter and exit the pin hole, and completes the unhooking action;

[0070] like Figure 7 As shown, the autonomous unhooking device for the unmanned tractor comprises a guide groove, a limit sensor, an unhooking motor, a pull rod and a traction pin.

[0071] The guide groove is designed as a guide device with a 45° inclined surface structure on the outer groove. When the hook is automatically reversed, the traction ring of the tractor is guided into the groove to physically contact the limit sensor. Because the tractor has backward power when reversing, the traction ring can fully contact the limit sensor.

[0072] The limit sensor switch is always in a normally open state during the guiding docking stage. When the unmanned tractor autonomously reverses the hook, the towing ring of the trailer continuously physically contacts the limit sensor switch until it is in a closed state, triggering a signal and outputting a continuous low-level signal.

[0073] The integrated controller determines whether the docking state is completed by detecting the high and low level changes of the limit sensor. When the vehicle is automatically hooked, a low level signal is detected, and the integrated controller sends a command to the unhooking motor through CAN to control the pull rod to lift the guide groove device and guide the insertion of the traction pin; when the vehicle is automatically unhooked, after stopping at the unhooking task point, the upper computer automatic driving system sends an unhooking command to control the unhooking motor to lower the pull rod, so that the traction pin and the guide groove device are separated, and the traction ring of the trailer is no longer bound by the traction pin, which means that the unhooking action is completed.

[0074] like Figure 8 As shown, after the unmanned tractor can realize autonomous positioning and path planning, and after the unmanned tractor and the trailer are equipped with autonomous detachment devices, the unmanned tractor performs integrated positioning of the trailer's traction ring according to the following steps:

[0075] S201: The perception system of the unmanned tractor receives a start signal of an automatic hooking task issued by the decision system after the tractor stops at a hooking stop point;

[0076] To ensure that the controller's computing resources are not consumed when other tasks are in progress, the perception system only detects when receiving the start of the automatic hook task, and the signal is intended to be sent by the decision system after the hook stop point stops;

[0077] S202: When the perception system receives the automatic hook task start signal, it starts the rear 16-line laser radar node and analyzes the laser radar data for use in the subsequent processing module;

[0078] S203: Coordinate conversion. To ensure that more lasers can scan the pallet truck, the laser radar needs to be installed sideways. However, when using the radar, the radar point needs to be converted to a horizontal plane first, and then the radar coordinates need to be converted according to the pre-calibrated parameters.

[0079] S204: Ground segmentation, setting a point cloud height threshold, filtering out the laser radar point cloud data below this height. This step is used to remove the ground point cloud;

[0080] S205: Pallet truck detection. Pallet truck detection intends to use the reflective stickers pasted on the trailer board to show high reflection intensity under the laser radar to detect and locate the trailer board. A vertical row of reflective stickers is pasted on both sides of the trailer board. The reflective intensity of the stickers in the point cloud is significantly different from that of other objects (in the point cloud, the reflective stickers appear purple) to detect both sides of the trailer board.

[0081] S206: Positioning of the pallet truck and sending of results. The positioning of the pallet truck mainly depends on the relative position relationship between the trailer part and the towing ring of the pallet truck detected in the previous step. According to the relative geometric position of the towing ring to the towing rod at the front end of the trailer, the position of the towing ring in the radar coordinate system is calculated, and through coordinate conversion, its GPS position information is obtained and sent to the decision-making system.

[0082] like Fig. 9 As shown, the left picture is the original point cloud map of the pallet truck detected by the laser radar, and the right picture is the point cloud map after filtering out the noise and optimization. The outer frame of the pallet truck is drawn according to the detected boundary, and the left and right places where the reflective stickers are pasted are considered to be the corner points of the pallet truck. According to the geometric position relationship, the position information of the detachable towing ring is calculated.

[0083] like Fig.10 As shown in the figure, the integrated controller determines whether the docking state is completed by detecting the high and low level changes of the limit sensor. After detecting that the docking state is completed, the integrated controller sends a command to the unhooking motor through CAN to control the pull rod to lift the guide groove device to guide the insertion of the traction pin. After the integrated controller detects that the motor rises to the highest position, it considers that the hooking action is completed, and then feeds back to the upper computer automatic driving system. After receiving the action of the hook completion, the automatic driving system controls the tractor to continue driving. The unhooking task is the opposite. After the unmanned tractor stops at the unhooking task point, the upper computer automatic driving system sends an unhooking command to control the unhooking motor to lower the pull rod, causing the traction pin and the guide groove device to disengage, and the traction ring of the trailer is no longer bound by the traction pin. After the upper computer automatic driving system receives the feedback signal that the motor has dropped to the lowest position, it considers that the unhooking action is completed and controls the tractor to continue driving.

[0084] Specifically, the autonomous hooking operation of the unmanned tractor comprises the following steps:

[0085] S301: The unmanned tractor automatically drives to the parking point in the hooking area and stops, and the automatic hooking task begins;

[0086] S302: Sending a hook start signal to the sensing system to perform positioning detection on the towing ring of the trailer;

[0087] S303: At the same time, the unhooking motor controls the pull rod to lower the guide groove device to the lowest point, so as to guide the traction pin to be separated from the guide groove device;

[0088] S304: Through steps S201-S206, the position point of the trailer towing ring detected by the perception system is sent to the decision system. After receiving the position of the detection point of the towing ring, the decision system plans a local reversing trajectory according to the current position point of the tractor and the positioning point of the trailer towing ring;

[0089] S305: If the detection signal of the position of the traction ring is not received for a long time, it is considered that the detection has failed, the hooking task has failed, and manual intervention is requested;

[0090] S306: The integrated controller detects a signal that the drive motor has dropped to the lowest position, and feeds it back to the automatic driving decision system, indicating that the tractor unhooking device has completed hooking preparation;

[0091] S307: If the signal indicating that the driving motor has dropped to the lowest position is not received for a long time, the hooking task fails and manual intervention is requested;

[0092] S308: After the identification and positioning information of the trailer towing ring and the planned reversing trajectory are completed and a signal that the drive motor has been lowered into position is received, the unmanned tractor is controlled to reverse at a low speed and automatically hook up;

[0093] S309: During the reverse hooking process, if the limit sensor has not been received, the vehicle continues to reverse until it reaches the limit position for safe reverse. If the limit sensor has not been received, the vehicle stops and an alarm is sounded, the hooking fails, and manual intervention is requested.

[0094] S310: If the traction ring of the pallet truck enters the guide groove during the reversing process, the physical resistance limit sensor is closed, and the integrated controller receives a low-level signal, the vehicle stops;

[0095] S311: In the parking state, the unhooking motor is controlled to control the pull rod to rise to the highest position;

[0096] S312: If the feedback rising position signal is not received, continue to send the motor rising command. If the number of times Cnt is sent is greater than 10 and there is still no effect, the hook fails and manual intervention is requested;

[0097] S313: On the contrary, after receiving the rising in place signal, it is considered that the automatic hooking is successful and the task is completed. The unmanned tractor continues to drive.

[0098] like Fig.11 As shown, the autonomous uncoupling operation of the unmanned tractor comprises the following specific steps:

[0099] S401: The unmanned tractor automatically drives to the designated uncoupling area and stops, and the automatic uncoupling task begins;

[0100] S402: The decision system issues a command to lower the unhooking motor, and controls the pull rod to lower the guide groove device to the lowest point;

[0101] S403: If the number of times Cnt of sending the unhooking motor descent command is less than 10, continue to wait for the integrated controller to detect the feedback descent in place signal. If the descent in place signal is not received, the automatic driving decision system determines that the unhooking has failed and requests manual intervention;

[0102] S404: If the automatic driving decision system receives a descent-in-place signal detected by the integrated controller, the automatic uncoupling is successful and the mission ends;

[0103] S405: Finally, a motor raising command is sent to lift the guide groove device into place, restore the normal state, and the unmanned tractor continues to travel.

[0104] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for an unmanned tractor to realize autonomous detachment indoors and outdoors, characterized in that: The method comprises: Step 1: Build a fusion framework autonomous positioning system based on factor graph optimization, calculate the real-time posture information of the unmanned tractor based on satellite position constraints, IMU pre-integration constraints, and lidar constraints to obtain real-time positioning information; Step 2: equip the unmanned tractor with an autonomous detaching device and equip the trailer with a traction ring, wherein the autonomous detaching device can identify the traction ring through a laser radar; Step 3: Based on the real-time positioning information of the unmanned tractor, plan the global path of the unmanned tractor and the local reversing path of the hooking area; Step 4: In combination with the real-time posture information and the planned local reversing path of the hooking area, the autonomous unhooking device performs autonomous hooking and unhooking operations with the towing ring.

2. The method for realizing autonomous indoor and outdoor detachment of an unmanned tractor according to claim 1, characterized in that: The step 1 includes taking the posture of the unmanned tractor as an optimization variable, and dividing it into unary edges and binary edges according to the number of optimization variables bound by constraints, wherein the satellite position constraint is a unary edge, and the IMU pre-integration and lidar odometer constraints are binary edges; and realizing high-precision autonomous positioning of the unmanned tractor by fusing observation information from multiple sensors.

3. The method for realizing autonomous indoor and outdoor detachment of an unmanned tractor according to claim 2, characterized in that: The lidar odometry constraints include: S101: Extract features from the point cloud generated by the laser radar and calculate the curvature of the point cloud: ， in, is the laser radar coordinate system Department Line Column of laser points, for The surrounding laser point collection; S102: performing weighted calculation and filtering on the curvature according to the distance to obtain a curvature effective value; S103: judging whether a point in the laser point cloud is a line feature point or a plane feature point according to the size of the effective value of the curvature; if the curvature value exceeds a set threshold, the corresponding point is a line feature point, otherwise it is a plane feature point; S104: Projecting the obtained line feature points and plane feature points onto a two-dimensional bird's-eye view; S105: Exploitation The optical flow method performs fast coarse registration of two-dimensional sparse point clouds to obtain matching pairs of feature points between two consecutive frames; S106: After obtaining the matched pairs of feature points, define the objective function of the rough registration: ， in, is the number of point clouds projected into the bird's-eye view, is the rotation matrix, is the translation matrix, and are the corresponding pairs of feature points between two frames obtained by the optical flow method, is the coarse registration transformation matrix; S107: Use the optical flow method to obtain a rough registration result, guide the three-dimensional sparse point cloud based on NDT to perform fine registration, and obtain the final LiDAR posture constraint result.

4. The method for realizing autonomous indoor and outdoor detachment of an unmanned tractor according to claim 1, characterized in that: The step 2 comprises: A decoupling device is arranged at the bottom of the rear end of the unmanned tractor, and a laser radar is arranged above the decoupling device; The end of the traction rod in the middle of the front end of the trailer that docks with the unhooking device of the unmanned tractor is provided with a trailer hanging point corresponding to the unmanned tractor, namely the traction ring; the left and right reflective strips are respectively provided on both sides of the front end of the trailer; The unmanned tractor screens the laser radar data according to the specific reflectivity to find the position of the reflective strip, and calculates the relative coordinate position of the traction ring relative to the tractor.

5. The method for realizing autonomous indoor and outdoor detachment of an unmanned tractor according to claim 4, characterized in that: The unmanned tractor screens the laser radar data according to the specific reflectivity to find the position of the reflective strip, and calculates the relative coordinate position of the traction ring relative to the tractor, and the specific steps include: S201: The perception system of the unmanned tractor receives a start signal of an automatic hooking task issued by the decision system after the tractor stops at a hooking stop point; S202: When the perception system receives the automatic hook task start signal, the rear 16-line laser radar node is started; S203: converting the laser radar data to a horizontal plane, and then converting the radar data coordinates according to pre-calibrated parameters; S204: setting a laser radar point cloud height threshold, filtering out laser radar point cloud data below the height threshold, so as to remove ground point clouds; S205: Utilizing the characteristic that the reflection intensity of the reflective strip in the point cloud is significantly different from that of other objects, the data is filtered to detect both sides of the trailer board; S206: Calculate the position of the towing ring in the radar coordinate system according to the relative geometric position of the towing ring to the towing rod at the front end of the trailer, obtain the GPS position information of the towing ring through coordinate conversion, and send it to the decision system.

6. The method for realizing autonomous indoor and outdoor detachment of an unmanned tractor according to claim 1, characterized in that: The step 3 comprises: Step 3.1, on the unmanned tractor-trailer-drivable road from the apron to the baggage carousel at the air cargo terminal at the airport, a global driving path of the unmanned tractor is planned based on the real-time positioning information to reach the stop point in the hook area; Step 3.2, the planning of the local reversing path in the hooking area includes: after the unmanned tractor stops at the stop point in the hooking area, the perception system detects and identifies the traction ring of the trailer, and sends the traction ring coordinate information to the decision system. The decision system plans a reversing path in a two-point linear interpolation method based on the positioning coordinates of the traction ring and the current coordinates of the vehicle to ensure that the traction ring of the trailer enters the guide groove of the tractor and triggers the limit switch signal after sufficient collision.

7. The method for realizing autonomous indoor and outdoor detachment of an unmanned tractor according to claim 1, characterized in that: The autonomous hooking operation performed by the autonomous unhooking device with the traction ring in step 4 includes: S301: The unmanned tractor automatically drives to the stop point in the hooking area and stops, and the automatic hooking task begins; S302: Sending a hook start signal to the sensing system to perform positioning detection on the towing ring of the trailer; S303: The unhooking motor controls the pull rod to lower the guide groove device to the lowest point, so as to guide the traction pin to be separated from the guide groove device; S304: Through steps S201-S206, the position point of the trailer towing ring detected by the perception system is sent to the decision system. After receiving the position of the towing ring detection point, the decision system plans a local reversing path according to the current position point of the towing vehicle and the positioning point of the trailer towing ring; S305: If the position detection signal of the traction ring is not received for a long time, it is considered that the detection has failed, the hooking task has failed, and manual intervention is requested; S306: The integrated controller detects a signal that the drive motor has dropped to the lowest position, and feeds it back to the decision system, indicating that the tractor unhooking device has completed hooking preparation; S307: If the signal indicating that the driving motor has dropped to the lowest position is not received for a long time, the hooking task fails and manual intervention is requested; S308: After completing the positioning information recognition of the trailer towing ring and the planned local reversing path, and receiving the signal that the drive motor has dropped to the lowest position, the unmanned tractor is controlled to reverse at a low speed and automatically hook up; S309: During the reverse hooking process, if the limit sensor has not been received, the vehicle continues to reverse until it reaches the limit position for safe reverse. If the limit sensor has not been received, the vehicle stops and an alarm is sounded, the hooking fails, and manual intervention is requested. S310: If the traction ring of the pallet truck enters the guide groove during the reversing process, the physical resistance limit sensor is closed, and the integrated controller receives a low-level signal, the vehicle stops; S311: In the parking state, the unhooking motor is controlled to control the pull rod to rise to the highest position; S312: If the feedback rising position signal is not received, continue to send the motor rising command. If the number of times Cnt is sent is greater than 10 and there is still no effect, the hook fails and manual intervention is requested; S313: If the rising in place signal is received, it is considered that the automatic hooking is successful and the task is completed.

8. The method for realizing autonomous indoor and outdoor detachment of an unmanned tractor according to claim 1, characterized in that: The autonomous unhooking operation performed by the autonomous unhooking device from the traction ring in step 4 includes: S401: The unmanned tractor automatically drives to the designated uncoupling area and stops, and the automatic uncoupling task begins; S402: The decision system issues a command to lower the unhooking motor, and controls the pull rod to lower the guide groove device to the lowest point; S403: If the number of times Cnt of sending the unhooking motor descent command is less than 10, continue to wait for the integrated controller to detect the feedback descent in place signal. If the descent in place signal is not received, the automatic driving decision system determines that the unhooking has failed and requests manual intervention. S404: If the automatic driving decision system receives a descent-in-place signal detected by the integrated controller, the automatic uncoupling is successful and the mission ends; S405: Send a motor raising command to lift the guide groove device into place, restore normal state, and the unmanned tractor continues to travel.

9. A system for autonomously unhooking a driverless tractor indoors or outdoors, characterized in that: include: The real-time positioning information unit is used to build a fusion framework autonomous positioning system based on factor graph optimization. Based on satellite position constraints, IMU pre-integration constraints, and lidar constraints, the real-time posture information of the unmanned tractor is calculated to obtain real-time positioning information; An assembly unit, used to assemble an autonomous detaching device for the unmanned tractor and a traction ring for the trailer, wherein the autonomous detaching device can identify the traction ring through a laser radar; A path planning unit, used to plan the global path of the unmanned tractor and the local reversing path of the hooking area based on the real-time positioning information of the unmanned tractor; The autonomous operation unit is used to combine the real-time posture information and the planned local reversing path of the hooking area. The autonomous unhooking device performs autonomous hooking and unhooking operations with the traction ring.

10. An electronic device, characterized in that: include: one or more processors; A memory for storing one or more programs; Among them, when one or more programs are executed by the one or more processors, the one or more processors implement the method for an unmanned tractor to achieve autonomous detachment indoors and outdoors as described in any one of claims 1-8.

11. A computer-readable storage medium, characterized in that: Executable instructions are stored thereon, and when the instructions are executed by the processor, the processor can implement the method for an unmanned tractor to achieve autonomous detachment indoors or outdoors as described in any one of claims 1-8.

Images