Path planning method and system of automatic transportation platform, storage medium and equipment

Abstract

The invention discloses a path planning method and system for an automatic transportation platform, a storage medium and equipment, and belongs to the technical field of path planning, and the method comprises the steps: collecting environment information through a multi-source heterogeneous sensor; target positioning is carried out based on environment information collected by a multi-source heterogeneous sensor, and target position information corresponding to different sensors is obtained; fusing the target position information of the multi-source heterogeneous sensor by using a Kalman filtering fusion algorithm to obtain a fused target state; and performing path planning based on the fused target state. According to the invention, through combination with the multi-source heterogeneous sensor, complementary sensing of environment information is realized; performing time synchronization and noise suppression on pilot frequency sensor data by using Kalman filtering and a weighted average algorithm to improve the target positioning precision; meanwhile, in combination with an improved path planning algorithm, an obstacle avoidance path is dynamically generated in a scene without a prior map.

Description

Technical Field

[0001] The present invention relates to the field of path planning technology, and in particular to a path planning method, system, storage medium and equipment for an automatic transport platform. Background Art

[0002] With the rapid development of automation technology, automatic transportation platforms are increasingly used in logistics, warehousing, intelligent manufacturing, public services and other fields. The key technologies for the application of automatic transportation platforms lie in target detection and path planning.

[0003] Traditional transportation platforms mostly rely on a single sensor (such as LiDAR or millimeter-wave radar) to achieve environmental perception and target tracking. In complex scenarios, they are easily limited by sensor performance, resulting in reduced tracking accuracy or even failure. For example, it is difficult for the visual system to stably detect targets in low-light or occluded scenes, and although millimeter-wave radar has the characteristics of strong penetration and high anti-interference ability, it lacks the ability to finely identify target categories. At present, the detection method of multi-sensor data fusion has been developed, such as collecting obstacle information through LiDAR combined with millimeter-wave radar, but the efficient fusion and synchronization of multi-source sensor data is still a major technical challenge, and the current multi-sensor data fusion method is mostly for homogeneous sensor data. When the sensor sampling frequencies vary greatly, traditional methods are difficult to ensure the temporal consistency and spatial matching of multi-source heterogeneous data.

[0004] Existing path planning algorithms usually rely on static maps or preset environmental models, which are not adaptable enough in dynamic environments and make it difficult to respond to sudden obstacles or changes in the target's motion trajectory in real time. Summary of the invention

[0005] The purpose of the present invention is to overcome the problems existing in the existing path planning methods, and to provide a path planning method, system, storage medium and device for an automatic transportation platform.

[0006] The objective of the present invention is achieved through the following technical solutions: In a first aspect, a path planning method for an automatic transport platform is provided, comprising: Use multi-source heterogeneous sensors to collect environmental information; Target positioning is performed based on the environmental information collected by multi-source heterogeneous sensors to obtain the target position information corresponding to different sensors; The target position information of multi-source heterogeneous sensors is fused using a Kalman filter fusion algorithm to obtain a fused target state; the target position information of multi-source heterogeneous sensors is fused using a Kalman filter fusion algorithm, including: Models are built based on the target position information corresponding to different sensors, and the measurement values ​​output by different models are weighted and fused; Path planning is performed based on the fused target state.

[0007] In some embodiments, the multi-source heterogeneous sensor includes a binocular camera and a millimeter-wave radar.

[0008] In some embodiments, the target positioning based on the environmental information collected by multi-source heterogeneous sensors includes: Use YOLOv11 to identify and locate targets in images captured by a stereo camera.

[0009] In some embodiments, the method further comprises: A buffer is used to store data from multi-source heterogeneous sensors within a certain period of time, and the Kalman filter synchronization algorithm is used to align and fuse multiple types of data streams.

[0010] In some embodiments, the weighted fusion of the measurement values ​​output by different models includes: Target state estimation after weighted fusion: ,in, R v Represents the covariance matrix corresponding to the binocular camera, R r Represents the covariance matrix corresponding to the millimeter-wave radar, tr ( R v )express R v trace, tr ( R r )express R r traces, Indicates the target state update corresponding to the binocular camera. Indicates the target status update corresponding to the millimeter-wave radar.

[0011] In some embodiments, performing path planning based on the fused target state includes: Environmental modeling and node division: Divide the current environment into grids, with each grid as a node, and establish a node graph with the current position of the automatic transport platform as the starting point and the target position as the end point; Cost function definition: ,in, represents the distance cost, represents the steering cost, represents the obstacle cost, α, β, γ are weight coefficients; Path search and optimization: add the starting point to the queue and record the shortest distance from the starting point to each node; use the greedy strategy to select the node with the smallest comprehensive cost in the queue as the current node and expand its adjacent nodes; when a new obstacle is detected, update the obstacle node status, recalculate the node cost of the affected area, and re-search the path from the current position; Simultaneous Localization and Mapping: Estimate the platform pose and build a map of the environment in real time.

[0012] In some embodiments, the real-time estimation of the platform posture and construction of the environment map includes: System state definition: define the platform posture state, environmental feature state and system state vector; Construct state transition model and observation model; Platform posture prediction and covariance prediction; Calculate algorithm gain and update platform posture state and covariance; When new feature points are detected, they are added to the environmental features, the system state vector is expanded and the covariance matrix is ​​updated to achieve incremental construction of the map.

[0013] In a second aspect, a path planning system for an automatic transport platform is provided, comprising: A data acquisition module, used to collect environmental information using multi-source heterogeneous sensors; The target recognition module is used to locate the target based on the environmental information collected by multi-source heterogeneous sensors and obtain the target position information corresponding to different sensors; The data fusion module is used to fuse the target position information of multi-source heterogeneous sensors using a Kalman filter fusion algorithm to obtain a fused target state; the use of the Kalman filter fusion algorithm to fuse the target position information of multi-source heterogeneous sensors includes: Models are built based on the target position information corresponding to different sensors, and the measurement values ​​output by different models are weighted and fused; A path planning module is used to perform path planning based on the fused target state.

[0014] In a third aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the path planning method for an automatic transport platform described in the first aspect is implemented.

[0015] In a fourth aspect, an electronic device is provided, comprising a memory and a processor, wherein the memory stores computer instructions executable on the processor, and when the processor executes the computer instructions, the path planning method for an automatic transport platform described in the first aspect is executed.

[0016] It should be further explained that the technical features corresponding to the above embodiments can be combined or replaced with each other to form a new technical solution without conflict.

[0017] Compared with the prior art, the present invention has the following beneficial effects: The present invention realizes complementary perception of environmental information by combining multi-source heterogeneous sensors; uses Kalman filtering and weighted average algorithm to perform time synchronization and noise suppression on heterogeneous frequency sensor data to ensure the temporal consistency and spatial matching of data and improve target positioning accuracy; at the same time, combined with an improved path planning algorithm, dynamically generates obstacle avoidance paths in scenarios without prior maps. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flow chart of a path planning method for an automatic transport platform shown in an embodiment of the present invention; Figure 2 A workflow diagram of a path planning system for an automatic transport platform shown in an embodiment of the present invention; Figure 3 A right bottom view of an intelligent transport vehicle shown in an embodiment of the present invention; Figure 4 A left side top view of an intelligent transport vehicle shown in an embodiment of the present invention; Figure 5 A wheel structure diagram of an intelligent transport vehicle shown in an embodiment of the present invention; In the figure: 1-right hand brake handle; 2-left hand brake handle; 3-horizontal rotator; 4-pitch rotator; 5-pitch rotator housing; 6-gyroscope; 7-material box; 8-power bracket; 9-automatic brake assembly; 10-battery compartment; 11-frame plate 11; 12-front wheel; 13-power output shaft; 14-left wheel; 15-left baffle of material box; 16-wheel lock switch; 17; manual lock; 18-manual lock tongue; 19-power connector; 20-wheel hub; 21-wheel hub connecting rod; 22-tire; 23-right camera; 24-millimeter wave radar; 25-right camera; 26-connecting rod; 27-main control board; 28-right baffle of material box; 29-right wheel; 30-right power motor; 31-left power motor; 32-quick release switch. DETAILED DESCRIPTION

[0019] The technical solution of the present invention is clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present application described and shown in the accompanying drawings can be arranged and designed in various configurations. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0020] It should be noted that the defects existing in the solutions in the above-mentioned prior art are the results obtained by the inventor after practice and careful research. Therefore, the discovery process of the above-mentioned problems and the solutions proposed in the embodiments of the present application for the above-mentioned problems below should all be the contributions made by the inventor to the present application in the process of invention and creation, and should not be understood as technical contents known to technical personnel in this field.

[0021] In view of the technical problems pointed out in the background technology, the embodiments provided by the present invention are as follows: Reference Figure 1 In an exemplary embodiment, a path planning method for an automatic transport platform is provided, comprising: Use multi-source heterogeneous sensors to collect environmental information; Target positioning is performed based on the environmental information collected by multi-source heterogeneous sensors to obtain the target position information corresponding to different sensors; The target position information of multi-source heterogeneous sensors is fused using a Kalman filter fusion algorithm to obtain a fused target state; the target position information of multi-source heterogeneous sensors is fused using a Kalman filter fusion algorithm, including: Models are built based on the target position information corresponding to different sensors, and the measurement values ​​output by different models are weighted and fused; Path planning is performed based on the fused target state.

[0022] Multi-source heterogeneous sensors refer to sensors of different types working together in the same application scenario. In this embodiment, binocular cameras and millimeter-wave radars are used as examples to illustrate the specific implementation process of path planning.

[0023] Among them, the binocular camera uses a target detection algorithm (such as YOLOv11) to process the images captured by the camera, identify and locate the targets in the image. The output includes the bounding box and category of the target. The target detection algorithm divides the image into multiple grids and predicts whether there is a target in each grid. Then it directly outputs the bounding box coordinates (x, y, w, h) and confidence. Then for each frame of the image, the center coordinates (x, y) of the target are extracted. Based on these coordinates, the direction and speed of the person's movement are calculated through the changes between consecutive frames. Specifically include: Motion tracking: The motion tracking algorithm uses the Kalman filter to track the position of the target in different frames. By calculating the changes between each frame, the direction and speed of the person's movement are obtained.

[0024] Speed ​​and direction calculation: The speed of the target is calculated based on the change of the target coordinates in consecutive image frames. The speed is obtained by dividing the displacement between two frames by the time interval: The specific calculation formula is as follows: .

[0025] The direction is obtained by calculating the coordinate change of the two frames before and after to get the angle. The method is: in two consecutive frames of images, the position coordinates of the target in the image (x 1 ,y 1 ) and (x 2 ,y 2 ), direction angle θ calculation formula:

[0026] This method can be used to calculate the direction angle between −180 degrees and 180 degrees.

[0027] In target tracking, millimeter wave radar accurately measures the target distance and estimates the target speed by transmitting and receiving millimeter waves. Specifically, it includes: Signal processing: The millimeter wave radar measures the time difference of the reflected electromagnetic waves, and calculates the distance of the target using the wave speed and reflection time. At the same time, the radar also calculates the relative speed of the target through the Doppler effect.

[0028] Target positioning: The reflected signal of the millimeter wave radar provides the distance between the target and the radar. Combined with multiple measurements at different angles, the position of the target can be obtained. The radar antenna in the present invention is an antenna array that can perform angle measurement.

[0029] Movement direction and speed: Millimeter-wave radar uses the Doppler effect to calculate the speed, and then obtains the target's movement trajectory and direction through continuous measurement.

[0030] The formula for speed calculation is: , where c is the speed of light, f 0 is the operating frequency of the radar, Δf is the frequency shift, and v is the velocity of the target.

[0031] Furthermore, since the binocular camera processes image data, the frame rate reaches 30 Hz to 60 Hz, while the millimeter-wave radar provides 10Hz to 20Hz data. In order to synchronize these data streams and improve accuracy, a buffer is used to store the binocular camera and millimeter-wave radar data for a certain period of time, and then the two types of data streams are aligned and fused through the Kalman filter synchronization algorithm.

[0032] The workflow of the fusion algorithm is: set the state equation of the system, and model the data of the binocular camera and the millimeter-wave radar separately, then perform weighted fusion on the measurement values ​​of the two sensors to finally obtain the optimal state estimate. The fusion algorithm has the ability to adaptively process noise, takes the measurement values ​​of vision and radar as observation input, and obtains the fused target state through prediction and update steps.

[0033] Specific implementation process of data fusion: 1. System state equation setting Assume that the state vector of the target in the two-dimensional plane is x=[x,y,vx,vy]T, where (x,y) is the target position and (vx,vy) is the target velocity. The state transfer equation is: x k =Ax k−1 +Bu k−1 +w k Among them, x k Represents the state variable at the current moment, x k−1 is the state variable at the previous moment, u k−1 is the input variable of the previous moment, A is the state transfer matrix, which describes how the system state is transferred from one moment to the next, △t is the time interval; B is the control input matrix, where it is assumed that the target moves at a uniform speed, u k−1 =0; w k is process noise, which obeys a Gaussian distribution with a mean of 0 and a covariance of Q. The state transfer matrix is ​​as follows: .

[0034] 2. Vision and radar data modeling Binocular camera: Output target center coordinates (xv, yv), the measurement equation is zv = H v x + V v , where V v is the visual measurement noise, the covariance matrix is ​​defined as , Represents the visual measurement matrix. The two matrices are as follows: , , where σ vx 2 and σ vy 2 They represent the variance of the visual sensor’s measurement noise in the x and y directions, respectively.

[0035] Millimeter-wave radar: Outputs target distance r and speed vr, and obtains target position (xr, yr) and speed (vrx, vry) through coordinate conversion. The measurement equation is zr = H r x + V r ,in is the radar measurement matrix, V r is the radar measurement noise, and the covariance matrix is ​​defined as .

[0036] 3. Weighted fusion and Kalman filtering steps Prediction steps:

[0037] in, It represents the predicted value of the state at time k based on all the information at time k - 1. represents the optimal estimate at time k - 1 based on all the information at time k - 1, Represents the covariance matrix of the state estimate at time k based on all the information at time k - 1. represents the covariance matrix of the optimal estimate based on all the information at time k - 1. Q represents the process noise covariance matrix, which describes the uncertainty in the system model.

[0038] Update steps: Calculate the Kalman gain for vision and radar: ,in, is the Kalman gain of the visual sensor (stereo camera) at time k, is the Kalman gain of the millimeter-wave radar at time k.

[0039] Calculate the state updates of vision and radar separately: ,in, and They are the state estimation values ​​based on the visual sensor and millimeter-wave radar at time k respectively. is the predicted state value, the same as in the state prediction step. represents the actual measurement value of the binocular camera at time k, Represents the actual measurement value of the millimeter-wave radar at time k.

[0040] Weighted fusion state estimation: ,in, Represents the final state estimate obtained by weighted fusion of the measurements of the visual sensor and the radar sensor at time k. tr ( R v )express R v trace, tr ( R r )express R r traces, Indicates the target state update corresponding to the binocular camera. Represents the target state update corresponding to the millimeter-wave radar. The trace of the covariance matrix is ​​used to measure the size of the sensor measurement noise, and the weight is inversely proportional to the noise.

[0041] Covariance update: ,in, represents the covariance matrix of the optimal estimate based on all the information at time k. I is the identity matrix, which is used to ensure the dimensional consistency of matrix operations. K k and H k are the fused gain matrix and measurement matrix respectively.

[0042] Aiming at the frequency difference between visual and radar data, the present invention adopts a buffer combined with a Kalman filter synchronization algorithm to achieve time alignment, introduces a weighted average algorithm, and dynamically allocates weights according to the visual pixel error and the radar distance error covariance to improve the fusion accuracy.

[0043] After obtaining the fused target state through the fusion method, path planning is performed. Specifically, the optimal path is calculated based on the path planning and obstacle avoidance algorithm. The core purpose of the path planning and obstacle avoidance algorithm is to calculate the optimal path based on the target position and obstacles in the environment. The algorithm optimizes the search process by the distance between the target position and the current position. The algorithm combines the breadth-first search and greedy algorithms, divides the path into multiple nodes, evaluates the cost of each node, and calculates the shortest path from the starting point to the end point without building a map in advance. In complex environments, the path planning algorithm that combines synchronous positioning and map building technology generates dynamic maps and path planning, and adjusts the path in real time to avoid new obstacles.

[0044] The specific implementation process of the path planning and obstacle avoidance algorithm is as follows: 1. Environment modeling and node division The current environment is divided into a grid, each grid is a node, and the node coordinates are (i, j), where i and j are grid indexes. The node status includes whether it is an obstacle, whether it is reachable, etc. The node graph is established with the current position of the automatic transport platform as the starting point and the target position as the end point.

[0045] 2. Cost function definition Distance cost: ,in, Represents the horizontal coordinate of the node on the two-dimensional plane, Represents the vertical coordinate of the node on the two-dimensional plane, represents the horizontal coordinate of the target point on the two-dimensional plane, Represents the ordinate of the target point on the two-dimensional plane, Represents the Euclidean distance from node n to the target location, guiding the path towards the target.

[0046] Obstacle cost: If node n is an obstacle, gobs(n) = ∞; otherwise gobs(n) = 0, ensuring that the path avoids obstacles.

[0047] Turning cost: gturn(n) =θprev,n, where θprev,n is the turning angle between the current node and the previous node, to avoid frequent turning of the path.

[0048] The comprehensive cost function is: , where α, β, and γ are weight coefficients, which are adjusted dynamically according to the environment.

[0049] 3. Path search and optimization Search initialization: add the starting point to the queue and record the shortest distance from the starting point to each node.

[0050] The greedy strategy selects the next node: select the node with the smallest comprehensive cost in the queue as the current node, and expand its adjacent nodes (up, down, left, right, upper left, lower left, upper right, lower right).

[0051] Real-time obstacle detection and path adjustment: When a new obstacle is detected, the obstacle node status is updated, the node cost of the affected area is recalculated, and the path search is restarted from the current position.

[0052] Path smoothing: Bezier curve fitting is performed on the generated path to reduce inflection points and make the transport platform run more smoothly.

[0053] 4. Combining simultaneous positioning and map construction This part is the core technology for the automatic transportation platform to achieve autonomous navigation in an environment without prior maps. By integrating visual and millimeter-wave radar data, it estimates the platform's posture in real time and builds an environmental map, providing dynamically updated high-precision environmental information for path planning. The following are the specific implementation details: 4.1. Definition of the status of the simultaneous positioning and mapping system Platform posture state: Assume that the posture vector of the platform in the two-dimensional plane is s = [x, y, θ] T , where (x, y) is the position of the platform in the global coordinate system (unit: meter); θ is the platform heading angle (unit: radian, defined as the angle between the platform axis and the positive direction of the x-axis).

[0054] Environmental feature state: Static obstacles and dynamic features in the environment are modeled as a feature point set m = [m 1 ,m 2 , m n ] T , each feature point m i = [x i , yi ] T ) represents its position in the global coordinate system.

[0055] System state vector: Combine the posture and environmental features to form an extended state vector: X = [S T , m T ] T .

[0056] 4.2. Implementation of the simultaneous positioning and mapping system State transition model: Assume that an automatic transport platform adopts wheel drive, including front wheels and driven rear wheels, and the driven rear wheels include left and right wheels. The platform motion adopts a differential drive model. Assume that the speeds of the left and right wheels are v l and v r (Unit: m / s), then within the time interval △t, the posture update equation is: , where L is the wheelbase of the left and right wheels (unit: meter); w x ,w y ,w θ is the process noise that obeys Gaussian distribution, and the covariance matrix is ​​Q.

[0057] Observation model: Visual sensor observation: The visual module detects environmental feature points (such as corners and landmarks) through YOLOv11, obtains the coordinates (u, v) of the feature points in the camera coordinate system after coordinate transformation, and converts them into the observation equation in the global coordinate system: .

[0058] Millimeter-wave radar observation: The radar detects the distance r and angle Φ of the target and converts them into the observation equation in the global coordinate system: .

[0059] The iterative steps of the algorithm are as follows: 1. Prediction stage Pose prediction:

[0060] Covariance prediction:

[0061] in, It represents the predicted value of the system state s at time k based on all the information at time k - 1\. f(.) is the state transition function, which describes how the system state transfers from one moment to the next. It represents the optimal estimate of the system state at time k - 1 based on all the information at time k - 1. Indicates the speed of the left wheel of the transport platform, Indicates the speed of the right wheel of the transport platform. is the Jacobian matrix of the state transfer matrix.

[0062] 2. Update phase Calculate the observation Jacobian matrix H v,k and H r,k ; Among them, H v,k Represents the visual sensor observation equation for the state vector X in The first-order partial derivative matrix at ; H r,k The millimeter wave radar observation equation represents the state vector X in The first-order partial derivative matrix at .

[0063] Algorithm Gains: ;in, is the Kalman gain matrix, which determines how much to trust the measurements and predictions when updating the state estimate. H v,k and / H r,k， Depends on which sensor is used for observation. is the measurement noise covariance matrix, which describes the uncertainty in the measurement process.

[0064] Status Update: ;in, Represents the predicted state value, which is consistent with the state prediction step Similarly, X here is the extended state vector, which contains information such as system state and environmental characteristics. According to the predicted state The calculated predicted values ​​of the measurements.

[0065] Covariance update: .

[0066] 3. Map Update When a new feature point is detected, it is added to the environment feature vector m, the state vector is expanded and the covariance matrix is ​​updated to achieve incremental construction of the map.

[0067] In another exemplary embodiment, based on the same inventive concept as the method embodiment, a path planning system for an automatic transport platform is provided, comprising: A data acquisition module, used to collect environmental information using multi-source heterogeneous sensors; The target recognition module is used to locate the target based on the environmental information collected by multi-source heterogeneous sensors and obtain the target position information corresponding to different sensors; The data fusion module is used to fuse the target position information of multi-source heterogeneous sensors using a Kalman filter fusion algorithm to obtain a fused target state; the use of the Kalman filter fusion algorithm to fuse the target position information of multi-source heterogeneous sensors includes: Models are built based on the target position information corresponding to different sensors, and the measurement values ​​output by different models are weighted and fused; A path planning module is used to perform path planning based on the fused target state.

[0068] Specifically, Figure 2 As shown in the figure, the data acquisition module includes a binocular camera and a 77G millimeter-wave radar. The binocular camera obtains environmental image information, and the 77G millimeter-wave radar completes tracking scenes in low-light environments or with many obstacles. The data collected by the binocular camera and the 77G millimeter-wave radar are sent to the central control unit. After receiving the data, the visual tracking module in the target recognition module: after the image data is processed, the coordinates and movement direction of the person can be obtained. Target detection and tracking uses the deep learning model YOLOv11 for target recognition and tracking to determine the target position. The millimeter-wave radar tracking module in the target recognition module transmits and receives reflected signals through radar signals to determine the position and movement direction of the person. Target positioning and speed estimation is based on the reflected signal of the millimeter-wave radar to estimate the target distance and speed, and provide real-time tracking data.

[0069] Furthermore, the data fusion module fuses the data from visual tracking and millimeter-wave radar to remove noise and improve tracking accuracy. According to the timeliness and accuracy of the two data sources, weighted averaging and Kalman filtering are performed to optimize the results.

[0070] The path planning module makes control decisions and path planning modules, and makes real-time decisions based on the results of data fusion. For example, how to adjust the moving path of the transport platform, avoid collisions, and ensure efficient tracking of targets. Finally, according to the instructions of the path planning module, the platform is driven to move through motors and wheels to track the target position.

[0071] Reference Figure 3-Figure 5 In another exemplary embodiment, the present application also provides an automatic transportation platform that is intelligently controlled by the above-mentioned path planning method. The platform is a small car, including: A vehicle body, on which a frame plate 11 and a material box 7 arranged on the frame plate 11 are provided; Front wheels and rear wheels arranged under the vehicle body, the rear wheels comprising a left wheel 14 and a right wheel 29 arranged symmetrically on the left and right; The power motors (mounted on the power bracket 8) are symmetrically arranged on the left and right sides, including a left power motor 31 and a right power motor 30, which are connected to the left wheel 14 and the right wheel 29 through the power output shaft 13 respectively; A power connector 19, a manual lock 17 and a wheel lock switch 16 are arranged on the rear wheel. The manual lock 17 includes a retractable manual lock tongue 18, which is used to extend in the automatic mode to embed into the power connector 19 to transmit the power of the power motor to the corresponding wheel, and retract in the manual mode to disconnect the power of the power motor; the wheel lock switch 16 is used to manually control the extension and retraction of the manual lock tongue 18; The handlebar is connected to the frame through a connecting rod 26, and the connecting rod 26 is provided with a horizontal rotator 3 and a pitch rotator 4; the handlebar is provided with a multi-source heterogeneous sensor; in the automatic mode, the horizontal rotator 3 and the pitch rotator 4 are used to adjust the angle of the multi-source heterogeneous sensor; in the semi-automatic mode, the horizontal rotator 3 is used to detect the angle of manual steering, and the pitch rotator 4 is used to adjust the forward speed; A manual brake assembly, comprising a left-hand brake handle 2 and a right-hand brake handle 1 arranged on the handlebar; An automatic brake assembly 9, arranged on the vehicle body; The main control board 27 is arranged on the vehicle body and is connected to the power motor, the horizontal rotator 3, the multi-source heterogeneous sensor and the automatic brake component 9.

[0072] In this example, the multi-source heterogeneous sensor includes a binocular camera and a millimeter wave radar 24. Figure 1 As shown, the binocular camera includes a left camera installed on the left handlebar and a right camera installed on the right handlebar. By setting up multi-source heterogeneous sensors, multi-type sensing tracking is performed on the target, reducing the dependence on a single sensor, and combining a certain sensor data fusion method to improve the target detection accuracy.

[0073] In this example, a gyroscope 6 is provided on the connecting rod 26, and the gyroscope 6 is connected to the main control board 27 via an I2C bus for real-time detection of the posture and angular velocity of the vehicle.

[0074] In this example, the automatic brake assembly 9 is an electromagnetic brake assembly, which is connected to the power circuit of the power motor (the battery is arranged in the battery compartment 10). The material box 7 is provided with a quick release switch 32, and the left and right sides of the material box 7 are fixed by baffles (the right baffle 28 of the material box and the left baffle 15 of the material box).

[0075] In this example, the front wheels are universal wheels to achieve any steering operation.

[0076] In this example, a symmetrical mechanical connection structure is used between the power motor and the rear wheel to ensure synchronization of left and right power outputs.

[0077] In this example, the pitch rotator 4 is provided with the pitch rotator housing 5 to provide certain protection for the pitch rotator 4 .

[0078] Furthermore, if Figure 3 As shown, any rear wheel structure includes a manual lock 17 (this example is X6), a manual lock tongue 18, a power connector 19, an inner wheel hub 20, a wheel hub connecting rod 21, and a tire 22. In semi-automatic and automatic modes, the manual lock tongue 18 pops out and connects the power connector 19. The power source comes from the power motor output. In manual mode, the manual lock tongue 18 retracts to disconnect the power, which can save people effort when pushing the trolley. The extension and retraction of this manual lock tongue 18 are manually activated by the wheel lock switch 16.

[0079] The left-hand brake handle 2 and the right-hand brake handle 1 have two functions, the first is the emergency brake function, and the second is the switching semi-automatic function. In the automatic mode, the car automatically tracks the target tracked by the radar and the camera. The horizontal rotator 3 and the pitch rotator 4 adjust different angles for real-time tracking. In the semi-automatic mode, the horizontal rotator 3 is used for the angle sensing function of turning left and right. The pitch rotator 4 is used to adjust the speed of the car. The faster you press down, the slower you lift up. At this time, the power motor has power output. The automatic brake component 9 can be activated in semi-automatic and automatic modes.

[0080] Furthermore, the left power motor 31 and the right power motor 30 work in coordination with various actuators through precise mechanical structures and control links, thereby ensuring flexible control of the automatic transport platform in different modes.

[0081] In the power transmission system, the left power motor 31 is connected to the inner wheel hub 20, the wheel hub connecting rod 21 and the tire 22 through the power output shaft 13 to drive the left wheel 14 to rotate, and the right power motor 30 is connected to the right wheel 29 in a symmetrical structure to drive the right wheel 29 to rotate. Both cooperate with the manual lock 17 through the power connector 19 to achieve the power switching of manual / automatic mode: in the automatic mode, the manual lock tongue 18 extends out and embeds into the power connector 19, and the power of the power motor is transmitted to the wheel through the transmission assembly; in the manual mode, the manual lock tongue 18 is retracted and disconnected, and the wheel can rotate freely to facilitate manual pushing.

[0082] In the execution part, the horizontal rotator 3 receives the instruction of the main control board 27 to adjust the steering angle and control the wheel steering in the automatic mode; in the semi-automatic mode, the horizontal rotator 3 acts as an angle sensor and transmits the angle signal of the manual steering to the main control board 27 to achieve accurate analysis and control of the steering angle. The pitch rotator 4 is connected to the pitch rotator housing 5 and the motor speed control module (set in the main control board 27). In the semi-automatic mode, the potentiometer resistance is adjusted to change the motor output power by pressing down or lifting up to achieve speed control, wherein pressing down accelerates (maximum 1.5m / s) and lifting up decelerates (minimum 0.3m / s), and excessive operation is prevented by the built-in limit block. The automatic brake component 9 adopts electromagnetic brakes and is connected to the left and right motor power supply circuits. In an emergency, the brake signal is triggered by the right brake handle 1 or the left brake handle 2 to cut off the power supply to achieve emergency braking; in the semi-automatic and automatic modes, the brake is smoothly braked according to the control instruction to ensure the safe stop of the platform.

[0083] The connection between the sensor and the motor ensures the intelligent control of the system: the gyroscope 6 detects the platform attitude and angular velocity in real time, and transmits the data to the main control board 27 through the I2C bus to adjust the left and right motor speeds and steering to maintain the balance of the platform; the wheel lock switch 16 manually controls the manual lock tongue 18 to switch the power connection. The manual mode feedback status is fed back to the main control board 27 through the motor speed and working current (when the speed is stable, the current is significantly smaller than the working current) so that the system can adjust the control strategy (path planning is disabled in manual mode, and sensor fusion is enabled in automatic mode).

[0084] Through the above-mentioned connection relationship, the left and right motors work together with the horizontal rotator 3, the pitch rotator 4, the brake assembly and other actuators, as well as the gyroscope 6, the wheel lock switch 16 and other sensors to form an efficient control system integrating power transmission, steering control, speed regulation, and safe braking, thereby realizing the reliable operation of the automatic transport platform in complex scenarios and flexible mode switching.

[0085] In another exemplary embodiment, based on the same inventive concept as the method embodiment, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the path planning method provided by the embodiment of the present invention is implemented. Based on such an understanding, the technical solution of this embodiment is essentially or partly contributed to the prior art or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0086] In another exemplary embodiment, based on the same inventive concept as the method embodiment, an electronic device is provided, including a memory and a processor, wherein the memory stores computer instructions executable on the processor, and the processor executes the path planning method provided by the embodiment of the present invention when executing the computer instructions.

[0087] The processor may be a single-core or multi-core central processing unit or a specific integrated circuit, or one or more integrated circuits configured to implement the present invention.

[0088] Embodiments of the subject matter and functional operations described in this specification may be implemented in: tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or a combination of one or more of them. Embodiments of the subject matter described in this specification may be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier to be executed by a data processing device or to control the operation of the data processing device. Alternatively or additionally, the program instructions may be encoded on an artificially generated propagated signal, such as a machine-generated electrical, optical, or electromagnetic signal, which is generated to encode and transmit information to a suitable receiver device for execution by the data processing device.

[0089] The processes and logic flows described in this specification can be performed by one or more programmable computers executing one or more computer programs to perform corresponding functions by operating on input data and generating output. The processes and logic flows can also be performed by special purpose logic circuits, such as FPGAs (field programmable gate arrays) or ASICs (application-specific integrated circuits), and the apparatus can also be implemented as special purpose logic circuits.

[0090] Processors suitable for executing computer programs include, for example, general and / or special microprocessors, or any other type of central processing unit. Typically, the central processing unit will receive instructions and data from a read-only memory and / or a random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, the computer will also include one or more large-capacity storage devices for storing data, such as magnetic disks, magneto-optical disks, or optical disks, or the computer will be operably coupled to this large-capacity storage device to receive data from it or to transmit data to it, or both. However, the computer does not necessarily have such a device. In addition, the computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, just to name a few.

[0091] It should be understood that each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0092] The above specific implementation methods are detailed descriptions of the present invention. It cannot be determined that the specific implementation methods of the present invention are limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions and substitutions can be made without departing from the concept of the present invention, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A path planning method for an automatic transport platform, characterized in that: include: Use multi-source heterogeneous sensors to collect environmental information; Target positioning is performed based on the environmental information collected by multi-source heterogeneous sensors to obtain the target position information corresponding to different sensors; Use Kalman filter fusion algorithm to fuse the target position information of multi-source heterogeneous sensors to obtain the fused target state; The method of fusing target position information of multi-source heterogeneous sensors using a Kalman filter fusion algorithm includes: Models are built based on the target position information corresponding to different sensors, and the measurement values ​​output by different models are weighted and fused; Path planning is performed based on the fused target state.

2. The path planning method for an automatic transport platform according to claim 1, characterized in that: The multi-source heterogeneous sensor includes a binocular camera and a millimeter-wave radar.

3. The path planning method for an automatic transport platform according to claim 2, characterized in that: The target positioning based on the environmental information collected by multi-source heterogeneous sensors includes: Use YOLOv11 to identify and locate targets in images captured by a stereo camera.

4. The path planning method for an automatic transport platform according to claim 1, characterized in that: The method further comprises: A buffer is used to store data from multi-source heterogeneous sensors, and a Kalman filter synchronization algorithm is used to align and fuse multiple types of data streams.

5. The path planning method for an automatic transport platform according to claim 2, characterized in that: The weighted fusion of the measurement values ​​output by different models includes: Target state estimation after weighted fusion: ,in, R v Represents the covariance matrix corresponding to the binocular camera, R r Represents the covariance matrix corresponding to the millimeter-wave radar, tr ( R v )express R v trace, tr ( R r )express R r traces, Indicates the target state update corresponding to the binocular camera. Indicates the target status update corresponding to the millimeter-wave radar.

6. The path planning method for an automatic transport platform according to claim 1, characterized in that: The performing path planning based on the fused target state includes: Environmental modeling and node division: Divide the current environment into grids, with each grid as a node, and establish a node graph with the current position of the automatic transport platform as the starting point and the target position as the end point; Cost function definition: ,in, represents the distance cost, represents the steering cost, represents the obstacle cost, α, β, γ are weight coefficients; Path search and optimization: add the starting point to the queue and record the shortest distance from the starting point to each node; use the greedy strategy to select the node with the smallest comprehensive cost in the queue as the current node and expand its adjacent nodes; when a new obstacle is detected, update the obstacle node status, recalculate the node cost of the affected area, and re-search the path from the current position; Simultaneous Localization and Mapping: Estimate the platform pose and build a map of the environment in real time.

7. A path planning method for an automatic transport platform according to claim 6, characterized in that: The real-time estimation of platform posture and construction of environment map includes: System state definition: define the platform posture state, environmental feature state and system state vector; Construct state transition model and observation model; Platform posture prediction and covariance prediction; Calculate algorithm gain and update platform posture state and covariance; When new feature points are detected, they are added to the environmental features, the system state vector is expanded and the covariance matrix is ​​updated to achieve incremental construction of the map.

8. A path planning system for an automatic transport platform, characterized in that: include: A data acquisition module, used to collect environmental information using multi-source heterogeneous sensors; The target recognition module is used to locate the target based on the environmental information collected by multi-source heterogeneous sensors and obtain the target position information corresponding to different sensors; A data fusion module is used to fuse the target position information of multi-source heterogeneous sensors using a Kalman filter fusion algorithm to obtain a fused target state; The method of fusing target position information of multi-source heterogeneous sensors using a Kalman filter fusion algorithm includes: Models are built based on the target position information corresponding to different sensors, and the measurement values ​​output by different models are weighted and fused; A path planning module is used to perform path planning based on the fused target state.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it implements a path planning method for an automatic transport platform as described in any one of claims 1-7.

10. An electronic device comprising a memory and a processor, wherein the memory stores computer instructions that can be executed on the processor, wherein: When the processor runs the computer instructions, it executes the path planning method for an automatic transportation platform described in any one of claims 1-7.

Images