Lawn mower path planning method based on process control

Abstract

The invention discloses a path planning method of a lawn mower based on process control, and relates to the technical field of automatic control, and the method comprises the steps: obtaining point cloud data and the spatial position of an obstacle, generating an occupation grid map and a lawn state grid map, and generating a preliminary path based on the occupation grid map and the lawn state grid map; based on the initial path and the initial state of the mower, generating a reference speed sequence and setting motion control parameters; in the execution process, the motion state and the surrounding environment of the mower are monitored in real time, the safety distance between the mower and an obstacle is calculated, in the execution process, the motion state and the surrounding environment of the mower are monitored in real time, the safety distance is calculated, a path risk index is obtained, path re-planning is conducted according to the risk index, and the path and motion control parameters of the mower are dynamically adjusted. By dynamically adjusting the path and the control parameters, the mowing efficiency is effectively improved, repeated mowing and area omission are avoided, and the flexibility and adaptability of the mowing machine are enhanced.

Description

Technical Field

[0001] This invention relates to the field of automation control technology, and in particular to a path planning method for a lawnmower based on process control. Background Technology

[0002] With the continuous development of intelligent agricultural technology, lawnmowers, as automated equipment, are increasingly widely used in farmland, lawns, and other environments. Traditional lawnmower path planning methods typically rely on simple preset paths or path search algorithms based on random strategies. While these methods can accomplish basic mowing tasks, they often lack the ability to adapt in real time to complex environments (such as obstacles, varying grass lengths, etc.) and the dynamic behavior of the lawnmower. To address this issue, recent research has focused on sensor-feedback-based path planning methods. These methods rely on devices such as LiDAR and vision sensors to acquire real-time environmental data and dynamically adjust the path through algorithms to cope with changes in obstacles, thus improving the system's flexibility and efficiency.

[0003] However, existing lawnmower path planning methods still have certain limitations in practical applications, especially in dealing with complex environments and real-time control. On the one hand, most existing methods rely on map-based or static path planning, ignoring the real-time motion status of the lawnmower and feedback from the environment, resulting in an inability to cope with sudden obstacles or path deviations during execution. On the other hand, when adjusting the lawnmower path, existing methods usually only perform simple path corrections or obstacle avoidance, without fully utilizing process control technology to optimize and adjust the motion status. This leads to lower mowing efficiency in areas with high-density obstacles or complex lawn conditions, and may even result in repeated mowing or missed areas. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a path planning method for lawnmowers based on process control to solve the problems of insufficient flexibility in path adjustment and inadequate real-time control optimization in existing technologies.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a path planning method for a lawnmower based on process control, which includes: scanning the lawnmower's operating area to obtain point cloud data of the environment and the spatial location of obstacles within the operating area, generating a real-time environmental map of the mowing area, and detecting obstacles.

[0008] Based on the acquired point cloud data and the spatial location of obstacles, an occupation raster map and a lawn state raster map are obtained, and a preliminary path for the mowing area is generated based on the occupation raster map and the lawn state raster map.

[0009] Based on the initial path and the initial state of the lawnmower, a reference velocity sequence is generated for each waypoint on the initial path, and the motion control parameters of the lawnmower are set.

[0010] During the lawnmower's execution path, the movement status of the lawnmower and the surrounding environment are monitored in real time, the safe distance between the lawnmower and obstacles is calculated, and path risk indicators are obtained. Based on the path risk indicators, the lawnmower path is replanned.

[0011] Based on the acquired real-time obstacle information and the lawnmower's motion feedback, the path and motion control parameters of the lawnmower are dynamically adjusted.

[0012] As a preferred embodiment of the path planning method for a lawnmower based on process control described in this invention, the generation of a real-time environmental map of the mowing area includes obtaining point cloud data of the environment where the lawnmower is located by performing environmental scanning using a lidar.

[0013] The system acquires environmental images via a camera and processes them using a deep convolutional neural network to determine the spatial location of obstacles.

[0014] Point cloud data and the spatial location of obstacles are fused using a Kalman filter to generate a real-time environmental map of the mowing area.

[0015] As a preferred embodiment of the path planning method for a lawnmower based on process control described in this invention, the obstacle detection refers to calculating an avoidance path in real time by comparing the current path of the lawnmower with the location of obstacles in the real-time environmental map of the mowing area through a path planning algorithm.

[0016] As a preferred embodiment of the path planning method for a lawnmower based on process control described in this invention, the step of obtaining the occupancy grid map and the lawn state grid map includes projecting point cloud data and the spatial position of obstacles onto a ground coordinate system to obtain a two-dimensional occupancy grid map.

[0017] The occupied grid is expanded to extend the obstacle boundaries to a safe distance, forming a no-entry zone.

[0018] The obtained grass height and density are used as lawn state variables and mapped to a raster of the same resolution to obtain a lawn state raster map.

[0019] Align the lawn status raster map with the occupied raster map according to the raster index.

[0020] As a preferred embodiment of the path planning method for a lawnmower based on process control described in this invention, the preliminary path for generating the mowing area includes calculating the required coverage intensity on each passable grid and determining the sweep line step distance based on the required coverage intensity.

[0021] Connect the ends of adjacent sweep lines in a reciprocating sequence, and the connecting segment adopts a combination trajectory of circular arc and straight line that satisfies the minimum turning radius constraint;

[0022] If the connecting section intersects with the restricted area, a detour connecting section is generated along the passable side of the restricted boundary, and the corresponding waypoint is updated;

[0023] Calculate the distance field based on the occupancy grid map;

[0024] Perform an iterative smoothing on the waypoint sequence. Without breaking through the restricted area and the safety expansion zone, move the curvature change points away from the obstacles in the opposite direction of the distance field gradient, and ensure that the spacing between adjacent points and the change in direction meet the lawnmower motion constraints. After smoothing, output the preliminary path.

[0025] As a preferred embodiment of the path planning method for a lawnmower based on process control described in this invention, the steps of generating a reference speed sequence for each waypoint on the initial path based on the initial path and the initial state of the lawnmower, and setting the motion control parameters of the lawnmower, are as follows:

[0026] The lawnmower obtains its initial state before starting to execute the initial path;

[0027] The arc length and curvature of each connecting segment of the preliminary path waypoint sequence are calculated sequentially, and the corresponding lawn load is obtained by sampling at the waypoints based on the required cover intensity.

[0028] Curvature and lawn load constraints are set for the lawnmower speed to generate the upper bound of the reference linear velocity for each waypoint and form a reference velocity sequence;

[0029] Discretize the upper bound of the reference linear velocity to a sequence of waypoints to obtain the reference velocity sequence;

[0030] Based on the initial state, calculate the lateral and heading deviations of the lawnmower relative to the initial path starting point, determine the target convergence time based on the magnitude of the lateral and heading deviations, and back-calculate the lateral control parameters based on the target convergence time.

[0031] Determine speed control parameters based on a reference speed sequence;

[0032] The motion control parameters of the lawnmower include lateral control parameters and speed control parameters.

[0033] As a preferred embodiment of the path planning method for a lawnmower based on process control described in this invention, the specific steps for calculating the safe distance between the lawnmower and obstacles and obtaining path risk indicators are as follows:

[0034] At the beginning of each control cycle, the lawnmower collects point cloud data, camera images, wheel speed odometer data, and inertial measurement data.

[0035] By fusing wheel speed odometer and inertial measurement data, the current planar position, heading angle, linear velocity and angular velocity of the lawnmower are obtained, and the current state is used as the starting state for the path update of this cycle. The lateral deviation and heading deviation of the lawnmower relative to the current mowing path are calculated.

[0036] Ground segmentation is performed on point cloud data, non-ground points are retained and clustered to obtain candidate obstacle outlines;

[0037] Target detection is performed on the camera images to obtain obstacle bounding boxes, and the obstacle bounding boxes are projected into two-dimensional contours in the ground coordinate system to obtain the obstacle set for the current period;

[0038] Assign a unique identifier and velocity estimate to each obstacle in the obstacle set for this period;

[0039] Write the set of obstacles for this cycle into the occupied grid map, calculate the safe distance from any position to the nearest obstacle from the obstacle boundary point, and sample the distance along the forward preview arc length of the current mowing path to obtain the path risk index.

[0040] As a preferred embodiment of the path planning method for a lawnmower based on process control described in this invention, the lawnmower path replanning based on path risk indicators includes checking each sampling point within the forward preview arc length within the same control cycle. If the distance obtained by substituting the path point coordinates into the safety distance field is less than or equal to zero, then the current path is determined to be infeasible in the sense of safety margin, and local replanning is immediately triggered.

[0041] The local replanning includes starting from the current estimated state and ending at the starting point of the next target coverage segment in the current coverage sequence, generating an obstacle bypass connection segment within the passable area occupying the grid map, and splicing it with the subsequent coverage segment of the original mowing path.

[0042] After splicing, the connecting segments are smoothed by curvature constraints to obtain the updated waypoint sequence;

[0043] The updated waypoint sequence, corresponding segment type information, and forward reference speed sequence are written into the tracking reference value of the next control cycle, enabling the lawnmower to update while executing.

[0044] As a preferred embodiment of the path planning method for a lawnmower based on process control described in this invention, the specific steps of dynamically adjusting the path and motion control parameters of the lawnmower based on the acquired real-time obstacle information and the motion feedback of the lawnmower are as follows:

[0045] Project the current state onto the current execution path and obtain the lateral deviation, heading deviation, velocity deviation, and curvature following error as motion feedback quantities;

[0046] The tracking deviations of path risk indicators and motion feedback are normalized on the same scale and then averaged integrally to obtain joint triggering indicators.

[0047] Within the same control cycle, the path and motion control parameters of the lawnmower are dynamically adjusted based on the triggering rule determination of the joint triggering index.

[0048] As a preferred embodiment of the path planning method for a lawnmower based on process control described in this invention, the triggering rule includes: if any point within the forward preview range satisfies the condition that the distance obtained by substituting the path point coordinates into the safety distance field is less than or equal to zero, then the current path is determined to be infeasible, and local replanning and parameter tuning are performed.

[0049] Otherwise, if the joint triggering index is greater than or equal to the replanning threshold, then perform local replanning and parameter tuning;

[0050] Otherwise, if the combined triggering index is greater than or equal to the parameter tuning threshold, parameter tuning will be performed.

[0051] Otherwise, keep the mowing path unchanged and only update the reference amount for the next cycle;

[0052] The parameter adjustment includes continuously reducing the upper limit of the reference speed and the steering gain parameter by a preset scaling factor in each control cycle.

[0053] The beneficial effects of this invention are as follows: By using sensor-based environmental scanning and dynamic path planning technology, the lawnmower can achieve real-time path adjustment and obstacle avoidance in complex environments, greatly improving the lawnmower's flexibility and adaptability; by combining process control technology to optimize motion control parameters, the speed and path can be dynamically adjusted during the mowing process to improve mowing efficiency and accuracy, effectively avoiding repeated mowing and missed areas. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a flowchart of a path planning method for a lawnmower based on process control.

[0056] Figure 2 This is a flowchart of the environmental scanning and obstacle detection process for the lawn mowing area.

[0057] Figure 3 A flowchart illustrating the process of generating a preliminary path for the mowing area.

[0058] Figure 4 This is a flowchart for replanning mowing routes based on path risk indicators.

[0059] Figure 5 This is a risk map showing the location of obstacles and the path.

[0060] Figure 6 This is a diagram showing the relationship between path risk and control parameters. Detailed Implementation

[0061] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0062] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0063] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0064] Reference Figures 1-6 This is one embodiment of the present invention, which provides a path planning method for a lawnmower based on process control, including the following steps:

[0065] S1. Perform an environmental scan of the lawnmower's operating area to obtain point cloud data of the environment and the spatial location of obstacles within the lawnmower's operating area, generate a real-time environmental map of the lawnmower's operating area, and perform obstacle detection.

[0066] Furthermore, the lawnmower uses LiDAR to scan the environment. LiDAR emits laser beams in all directions and receives the return signals. By calculating the return time of the signals, it obtains the distance information of surrounding objects. The data generated by LiDAR is a point cloud dataset, which represents the position of each point in the surrounding environment. The point cloud data of each scan represents the distance information of different positions around the lawnmower, represented by the spatial coordinates of each point.

[0067] Furthermore, the lawnmower's camera captures environmental images and uses computer vision algorithms to identify obstacles in the images in real time. Static obstacles are identified through image processing algorithms, and the position and distance of the objects in the image are calculated through grayscale and depth analysis, thereby generating obstacle boundary information. Among these, obstacle detection can be processed by a deep convolutional neural network (CNN), which takes the environmental images captured by the camera as input and outputs the spatial position of the obstacles.

[0068] The deep convolutional neural network (DCNN) takes normalized RGB images from a camera as input. It first extracts three-scale features through a backbone network consisting of multiple layers of Conv-BN-SiLU and residual bottleneck stacking, then performs multi-scale feature fusion through top-down upsampling and stitching using FPN and bottom-up downsampling and stitching using PAN. Finally, at each scale, a decoupled detection head is set to output bounding box regression values, obstacle confidence, and class probabilities to form obstacle location results. The training steps involve collecting and labeling a dataset of lawn mowing scene images, performing data augmentation and batch training at a fixed input size, using the average arithmetic sum of bounding box regression loss, confidence loss, and classification loss as the objective function, iteratively updating network parameters using backpropagation, and then fixing the deep convolutional neural network parameters on a validation set.

[0069] Furthermore, the point cloud data from the LiDAR and the obstacle recognition results from the camera are fused to generate a real-time environmental map of the lawnmower. Specifically, the data fusion is optimized and noise is removed using a Kalman filter. By setting the time synchronization parameters of the LiDAR and the camera, the final fused data is obtained through the Kalman filter, and a real-time environmental map is generated.

[0070] Furthermore, the location of obstacles is identified and marked using the fused real-time environmental map. Specifically, obstacle detection calculates the avoidance path in real time by comparing the current path of the lawnmower with the location of obstacles in the environmental map. The avoidance path is planned in the environmental map by a path planning algorithm, such as the A* algorithm, and the movement trajectory of the lawnmower is dynamically adjusted to avoid collisions with obstacles.

[0071] S2. Based on the acquired point cloud data and the spatial location of obstacles, obtain the occupancy grid map and the lawn state grid map, and generate the preliminary path of the mowing area based on the occupancy grid map and the lawn state grid map.

[0072] Furthermore, after completing environmental scanning and obtaining a real-time environmental map, the onboard processor generates a passable grid map and a lawn status grid map in the same coordinate system. Based on the two grid maps, a preliminary path that can be directly executed is formed. Specifically, the LiDAR point cloud and camera recognition results are projected onto the ground coordinate system to obtain a two-dimensional occupied grid map. The occupied grid map is then expanded to extend the obstacle boundary to a safe distance to form a no-entry zone. At the same time, the grass height and density obtained by the camera are mapped as lawn status variables to a grid of the same resolution to obtain a lawn status grid map. The lawn status grid map is then aligned with the occupied grid map one by one according to the grid index.

[0073] Furthermore, in order to automatically densify the coverage of the initial path in areas with higher or denser grass, the required coverage intensity is calculated on each passable grid, and the sweep line step distance, i.e. the spacing between adjacent travel trajectories, is determined by the required coverage intensity.

[0074] Demand coverage intensity is expressed as:

[0075] ;

[0076] in, Indicates the strength of demand coverage. Represents the horizontal coordinate in the ground coordinate system. Represents the vertical coordinates of the ground coordinate system. Indicates the influence coefficient of grass height. Indicates the actual grass height. Indicates the height of the benchmark grass. Represents a very small positive number. Indicates the influence coefficient of lawn density. Indicates the actual lawn density. This indicates the upper limit for observing lawn density.

[0077] It should be noted that the grass height influence coefficient is a calibration constant obtained by repeatedly testing the mowing effect of the lawnmower in different grass height areas and optimizing the path and efficiency of the lawnmower for different grass heights; the lawn density influence coefficient is calibrated by analyzing the impact of lawn density on the mowing efficiency of the lawnmower and combining it with lawn type data in the actual mowing process.

[0078] The sweep line step distance is expressed as:

[0079] ;

[0080] in, Indicates the sweep line step distance. This indicates the effective cutting width reference value.

[0081] It should be noted that, , In areas with taller grass, the required coverage intensity is greater, resulting in smaller sweep line spacing and a denser executable coverage area.

[0082] Furthermore, a restricted area and a safety extension zone for the restricted area are set up. The restricted area is topologically segmented, and the passable area is decomposed into several sub-regions with no or few holes. For each sub-region, a set of boundary points is extracted and the main direction of the sub-region is calculated. For example, the main eigenvector of the boundary point covariance matrix is ​​used as the main axis direction, and the main axis direction is used as the sweep direction to reduce the number of turns and reduce redundant coverage. Within each sub-region, a set of parallel sweep lines are generated along the sweep direction. The adjacent spacing of each sweep line is not constant, but is averaged within the local range of the sweep line based on the calculated sweep line step distance, and upper and lower limit clamping is applied to avoid sweeping too densely or too sparsely. The upper and lower limit clamping refers to limiting the spacing of each sweep line so that the spacing of each sweep line is between the calculated minimum and maximum spacing. Then, each sweep line is sampled as a waypoint sequence with a fixed arc length, and waypoints falling into the restricted area or the safety extension zone of the restricted area are removed. The arc length refers to the distance from one point to another along the path curve segment of the lawnmower.

[0083] Specifically, the determination of the restricted area involves the lawnmower synchronously acquiring LiDAR point clouds and camera obstacle bounding boxes in each control cycle. These obstacle bounding boxes are then mapped to a ground coordinate system to obtain an obstacle point set. The obstacle point set is then segmented on the ground, retaining non-ground points. Euclidean clustering is then performed on the non-ground points to obtain several obstacle clusters. A two-dimensional convex hull is calculated for each obstacle cluster to obtain a set of obstacle geometric boundary polygons. Each obstacle geometric boundary polygon is directly defined as a physical restricted area core. Simultaneously, non-mowing core areas are superimposed as additional restricted areas cores, such as tree roots, flower bed boundaries, step edges, and the inner side of fences. These additional restricted areas cores are semantically segmented and labeled by the camera, projected onto the ground coordinate system, and their outer boundary polygons are obtained. Finally, the union of all restricted areas cores is taken to obtain the restricted area.

[0084] The determination of the safe expansion width specifically involves calculating an expansion band width for each sensing cycle, which is then used to expand the restricted area outwards. The safe expansion width per sensing cycle is determined by the lawnmower vehicle size, positioning error, control tracking error, and sensing error. The safe expansion width is expressed as:

[0085] ;

[0086] in, Indicates the width of the safety strip extending outwards from the restricted area. This represents the equivalent external radius of the lawnmower. Indicates the current moment. Indicates the statistical time window. Indicates lateral tracking error. Indicates the time the lawnmower is at. The actual control input, This represents the normalization constant of the control input. Indicates the standard deviation of planar positioning. Indicates the error scaling coefficient. This represents the control scaling factor.

[0087] It should be noted that a two-dimensional envelope of the lawnmower's outer contour and the outer edge of the cutter head is established, and the maximum distance from the centroid projection point to the envelope boundary is taken as the equivalent outer radius. At the current moment, the lawnmower's centroid position and heading angle are obtained through fusion positioning. In the reference trajectory, i.e., the waypoint sequence, the reference point closest to the centroid position is found, and the trajectory tangential direction at the reference point is obtained. The projection length of the displacement vector from the centroid to the nearest reference point in the normal direction of the reference point is taken as the lateral tracking error. The drive command output by the controller is defined as the actual control input. The lawnmower uses odometer, IMU, and laser vision odometer for fusion positioning, and fusion filtering, such as extended Kalman filtering, is used to output the covariance matrix of the position estimate. The combination of the variances of the position estimate covariance matrix in the two directions of the planar position is taken as the planar positioning standard deviation.

[0088] It should be noted that the error scaling coefficient is obtained by least-squares fitting of historical lateral tracking errors and the corresponding actual safety distance increment required for expansion; the control scaling coefficient is obtained by regression fitting of the equivalent time of the control input and the actual additional safety distance increment within the same time window, based on the experience of the maximum lateral displacement caused by changes in control input.

[0089] The restricted area is morphologically expanded by the safe expansion width to obtain the expanded restricted set; then, a point inclusion determination is performed on the center point of each grid on the grid map; the point inclusion determination means that if the center of the grid falls into the restricted area, it is marked as a "restricted grid"; if it falls into the restricted set, it is marked as a "safe expansion zone grid"; otherwise, it is marked as a "passable grid".

[0090] Furthermore, adjacent sweep line ends are connected in a round-trip sequence, and the connecting segment adopts a circular arc-straight line combination trajectory that satisfies the minimum turning radius constraint; if the connecting segment intersects with the restricted area, a detour connecting segment is generated along the passable side of the restricted boundary, and the corresponding waypoint is updated.

[0091] The distance field is calculated based on the occupancy grid map. Specifically, for any point... Calculate the Euclidean distance to the nearest obstacle boundary; perform an iterative smoothing on the waypoint sequence, and without breaking through the restricted area and the safety expansion zone, slightly shift the curvature change points in the opposite direction of the distance field gradient, i.e. away from the obstacle, while ensuring that the spacing and turning changes of adjacent points satisfy the lawnmower motion constraints; output the preliminary path after smoothing.

[0092] Among them, lawnmower motion constraints refer to physical limitations such as turning radius, speed and acceleration of the lawnmower when executing a path, to ensure that the lawnmower can move safely and smoothly under given environmental conditions, so as to avoid exceeding the lawnmower's design capabilities; the preliminary path includes waypoint sequence and segment type information.

[0093] S3. Based on the initial path and the initial state of the lawnmower, generate a reference speed sequence for each waypoint on the initial path and set the motion control parameters of the lawnmower.

[0094] Furthermore, before starting to execute the initial path, the lawnmower reads the output of the positioning and attitude sensors to obtain the initial state, including the initial planar position, initial heading angle, initial linear velocity, and initial angular velocity; at the same time, it reads the discrete time reference used for calculation during the control cycle.

[0095] The arc length and curvature of each segment of the waypoint sequence of the preliminary path are calculated sequentially, and the corresponding lawn load is obtained by sampling at the waypoints based on the required coverage intensity. The lawn load refers to the additional power, torque or energy consumption required by the lawn mower during the cutting and driving process caused by the grass height and grass density factors when executing the path.

[0096] Curvature and lawn load constraints are set for the lawnmower speed to generate an upper bound for the reference linear velocity at each waypoint, forming a reference velocity sequence, represented as:

[0097] ;

[0098] in, Indicates the arc length along the path The upper limit of the reference linear velocity at that point, Indicates the path arc length. Indicates the base driving speed. Represents the coupling coefficient. This represents the curvature obtained from waypoint geometry calculations. Indicates the arc length The load on the lawn at that location.

[0099] It should be noted that the lawn load is obtained by sampling the grass height and density estimated by the camera at the waypoints, and by weighted normalization calculation based on the increment of the mower motor current relative to the no-load baseline.

[0100] It should be noted that the coupling coefficient is determined by conducting closed-loop tests under different curvature conditions, adjusting the relationship between the lawnmower's speed and the path curvature, until overspeeding is avoided during large path bends, thus preventing increased deviation or loss of control.

[0101] Discretize the upper bound of the reference linear velocity to the waypoint sequence to obtain the reference velocity sequence, and perform a limited climb on the reference velocity at the initial path start point based on the initial velocity.

[0102] Furthermore, the lateral and heading deviations of the lawnmower's initial state relative to the initial path starting point are calculated. The target convergence time is determined based on the magnitude of the lateral and heading deviations, and the lateral control parameters are calculated inversely based on the target convergence time. In order to achieve the requirement that the larger the deviation, the smoother the control; and the smaller the deviation, the tighter the control, the lateral weight parameters and steering gain parameters are set.

[0103] The steering gain parameter is expressed as:

[0104] ;

[0105] in, This represents the steering gain parameter corresponding to the heading error. Indicates the reference steering gain. Indicates the gain smoothing factor. Indicates lateral deviation. Indicates heading deviation. This represents the normalized scale of the lateral deviation. This represents the normalized scale of the heading deviation.

[0106] The horizontal weighting parameter is expressed as:

[0107] ;

[0108] in, This represents the lateral deviation weighting parameter. Indicates the baseline horizontal weights, This represents the weighting softening coefficient.

[0109] It should be noted that, on a flat, low-load, and unobstructed standard lawn, tracking is performed along a straight line and a fixed-curvature arc at the rated reference speed. The maximum available gain that prevents continuous oscillations in the steady-state heading error is used as the reference steering gain. Repeated starts are performed under different combinations of initial lateral deviations and heading deviations, with the maximum steering angular velocity not exceeding the limit as a constraint. The gain softening coefficient is determined by performing a single-parameter sweep on the gain softening coefficient and taking the minimum value that satisfies the constraint. On the standard lawn, with the target coverage accuracy as a constraint, the weight that optimizes the error index is selected as the reference lateral weight to minimize the average lateral error of the planned path tracking and ensure that the steering change rate does not exceed the limit. Path entry is performed under different initial lateral deviation conditions, with the lateral error convergence time and the maximum curvature command not exceeding the limit as constraints. The weight softening coefficient is then subjected to a single-parameter sweep, and the value that simultaneously satisfies the constraint and has the shortest convergence time is taken as the weight softening coefficient.

[0110] Speed ​​control parameters are determined based on a reference speed sequence, including the proportional and integral coefficients for speed tracking, and the upper and lower bounds of acceleration and angular velocity are calculated based on the motor's output torque, tire-ground adhesion conditions, and the upper limit of lawn load.

[0111] The reference velocity sequence, steering gain parameters, lateral weight parameters, acceleration upper and lower bounds, and angular velocity upper and lower bounds are all written into the parameter table corresponding to the control cycle.

[0112] During the initial control cycle, the rate of increase of the reference speed and the rate of change of the steering gain are locked to ensure that the changes in control parameters meet the constraints. If the initial state and the path start point are not found to meet the reachability requirements, the path start point neighborhood is entered at low speed along the shortest feasible arc, and then the lock is released to enter normal tracking.

[0113] S4. During the lawnmower's execution path, monitor the lawnmower's movement status and surrounding environment in real time, calculate the safe distance between the lawnmower and obstacles, obtain path risk indicators, and replan the lawnmower path based on the path risk indicators.

[0114] Furthermore, at the beginning of each control cycle, the lawnmower synchronously collects lidar point cloud, camera images, wheel speed odometer, and inertial measurement data according to the timestamp, and performs noise reduction and distortion compensation on the point cloud and images respectively, so that the data all fall into the same ground coordinate system.

[0115] By fusing wheel speed odometer and inertial measurement data, the current planar position, heading angle, linear velocity, and angular velocity of the lawnmower are obtained, and the current state is used as the starting state for the path update of this cycle; at the same time, the lateral deviation and heading deviation of the lawnmower relative to the current mowing path are calculated.

[0116] Ground segmentation is performed on the LiDAR point cloud, non-ground points are retained and clustered to obtain candidate obstacle contours; target detection is performed on the camera image to obtain obstacle bounding boxes, and the obstacle bounding boxes are projected into two-dimensional contours in the ground coordinate system; then spatial consistency matching is performed on the two types of contours to output the obstacle set for the current period, and obstacles in the obstacle set for the current period are assigned a unique identifier and velocity estimate to distinguish between newly added obstacles and existing obstacles.

[0117] Write the set of obstacles for the current cycle into the occupied grid map, and calculate the safe distance from any position to the nearest obstacle from the obstacle boundary point; then sample the distance along the forward preview arc length of the current mowing path to obtain the path risk index, which is used to determine whether the mowing path needs to be updated.

[0118] The safe distance is expressed as:

[0119] ;

[0120] in, Indicates the first Control cycle in ground coordinates Maintain a safe distance. Indicates the boundary points of the obstacle. Indicates the first The number of obstacle boundary points involved in distance calculation during the control cycle. Indicates the first Control cycle number The lateral coordinates of the boundary points of the obstacles Indicates the first Control cycle number The longitudinal coordinates of the boundary points of the obstacle.

[0121] Path risk indicators are expressed as follows:

[0122] ;

[0123] in, Indicates the first Path risk indicators for the control cycle Indicates the forward preview arc length. Indicates the first Control period in arc length The horizontal coordinate of the path at that location. Indicates the first Control period in arc length The vertical coordinate of the path at that location. This represents the distance obtained by substituting the path point coordinates into the safety distance field. This represents the integral operator.

[0124] Within the same control cycle, each sampling point within the forward preview arc length is checked for existence. If such a path exists, it is determined that the current path is infeasible in the sense of safety margin, and local replanning is immediately triggered.

[0125] Furthermore, when local replanning is triggered, a bypass connection segment is generated within the passable area of ​​the grid map, starting from the current estimated state and ending at the starting point of the next target coverage segment in the current coverage sequence. This segment is then spliced ​​with the subsequent coverage segment of the original mowing path. After splicing, the connection segment is smoothed by curvature constraints to obtain the updated waypoint sequence. Finally, the updated waypoint sequence, corresponding segment type information, and forward reference velocity sequence are written into the tracking reference value of the next control cycle, enabling the lawnmower to update while executing.

[0126] In this embodiment, to verify the lawnmower's path planning capability in complex environments, sensors such as LiDAR and cameras are used to enable the lawnmower to detect obstacles in the surrounding environment in real time and dynamically adjust its path based on the location of the obstacles; for example... Figure 5 As shown in the diagram, the path risk index reflects whether the lawnmower will collide with obstacles or perform tasks in dangerous areas; the dark red areas represent high-risk areas where the lawnmower needs to perform obstacle avoidance maneuvers. Based on real-time feedback, the lawnmower uses algorithms to calculate the optimal path to avoid these high-risk areas, thereby improving work efficiency and ensuring safety. Furthermore, the obstacle distribution in the diagram represents static and dynamic obstacles that the lawnmower needs to avoid, such as trees, rocks, pedestrians, or pets. The presence of these obstacles affects the lawnmower's movement trajectory and path planning.

[0127] S5. Based on the real-time obstacle information and the motion feedback of the lawnmower, dynamically adjust the path and motion control parameters of the lawnmower.

[0128] Furthermore, at the beginning of each control cycle, the obstacle set and safety distance output from the previous cycle are read, and the current state of the lawnmower is obtained synchronously; the current state is projected onto the current execution path, and the lateral deviation and heading deviation are obtained as motion feedback quantities.

[0129] By normalizing the path risk indicator and tracking deviation on the same scale and performing an integral average, a joint trigger indicator is obtained, expressed as:

[0130] ;

[0131] in, Indicates the first Joint triggering indicators for control cycles, Indicates the first Control cycle lawnmower predicted trajectory in arc length Lateral deviation at the location, This indicates the normalized scale of the predicted lateral bias. Indicates the first Control cycle lawnmower predicted trajectory in arc length The heading deviation at that point This indicates the normalized scale of the predicted heading deviation. This represents the lateral deviation weighting coefficient. This represents the heading deviation weighting coefficient.

[0132] It should be noted that, when conducting actual vehicle tracking under the standard lawn and representative obstacle layout, the goal is to minimize the root mean square of the lateral deviation within the forward preview interval, and the constraint is that the maximum angular velocity and the maximum rate of change of steering do not exceed the limits. A single-parameter scan is performed on the lateral deviation weight coefficient, and the minimum value that first satisfies the constraint and the target value no longer decreases is taken as the lateral deviation weight coefficient. The value range is generally [0.1, 10]. When conducting actual vehicle tracking on typical paths containing straight lines, arcs, and S-curves, the goal is to minimize the root mean square of the heading deviation, and the constraint is that the peak value of the lateral deviation and the number of trajectory oscillations do not exceed the limits. A single-parameter scan is performed on the heading deviation weight coefficient, and the value at the inflection point that reduces the heading error without introducing lateral overshoot is taken as the heading deviation weight. The value range is generally [0.1, 1].

[0133] It should also be noted that the predicted trajectory of the lawnmower refers to the forward prediction calculation used by the lawnmower for closed-loop tracking within the control cycle. Specifically, at the beginning of the control cycle, the lawnmower's controller uses the current state, i.e., position, heading, and speed, as initial values, takes the displacement sequence of the control variables executed in the previous cycle as the initial control sequence, and then gradually extrapolates in the preview time domain according to the discrete motion state of the lawnmower to obtain a series of predicted state points, i.e., predicted position and predicted heading. The trajectory formed by connecting the predicted state points is the predicted trajectory of the lawnmower.

[0134] Furthermore, within the same control cycle, the following A1-A4 execution steps are performed based on the joint triggering indicator triggering rules:

[0135] A1. If there exists any point within the forward preview range that satisfies... If the current path is deemed infeasible, local replanning and parameter tuning will be performed.

[0136] A2, otherwise if Then, local reprogramming and parameter tuning are performed; among them, The threshold for replanning.

[0137] A3, otherwise if Then, parameter tuning will be performed; among them, This is the threshold for parameter tuning.

[0138] A4. Otherwise, keep the path unchanged and only update the reference value for the next cycle.

[0139] It should be noted that, under standard lawn and typical obstacle layouts, the maximum value of the joint index that, upon triggering, ensures the minimum safe distance within the forward preview range is just above zero and that no path infeasibility subsequently occurs is calculated, and this maximum value is taken as the replanning threshold. The upper quantile value (e.g., the 95th quantile) of the joint index that can restore the tracking deviation to the allowable range without triggering replanning simply by reducing the reference speed and steering gain is also calculated, and this upper quantile value is taken while ensuring... As a parameter tuning threshold.

[0140] Furthermore, parameter tuning specifically involves continuously reducing the upper limit of the reference speed and the steering gain parameters by a preset scaling factor in each control cycle. The scaling factor is determined by performing a single-parameter sweep of different coefficient values ​​under standard lawn and typical obstacle scenarios, with the constraints and objectives being that the maximum deceleration, maximum angular velocity, and maximum deviation do not exceed the limits and the combined trigger index decreases the fastest. The minimum coefficient that satisfies the constraints is selected as the scaling factor.

[0141] The updated mowing path and the updated set of control parameters are written into the next control cycle; the next control cycle performs closed-loop tracking and repeats according to the updated reference path and reference speed.

[0142] In this embodiment, to verify the impact of dynamic adjustment of control parameters on obstacle avoidance and efficiency improvement, the lawnmower's trajectory and path are optimized by adjusting control parameters such as lateral control, speed control, and steering gain. Figure 6 As shown in the figure, the relationship between path risk indicators and lawnmower control parameters is illustrated. Each point represents the correspondence between lawnmower control parameters, including lateral control, speed control, and steering gain, and path risk within one experimental cycle. Figure 6 The blue, green, and orange colors in the diagram represent the relationship between lateral control, speed control, and steering gain and path risk, respectively. Figure 6 The invention verifies how a lawnmower can avoid collisions in high-risk areas by reducing speed or adjusting steering gain, while ensuring path optimization, thus improving mowing efficiency and accuracy. The dynamic adjustment of control parameters verifies the effectiveness of optimizing motion control parameters through process control in the invention, avoiding repeated mowing and missed areas.

[0143] In summary, this invention, through sensor-based environmental scanning and dynamic path planning technology, enables lawnmowers to adjust their paths and avoid obstacles in complex environments in real time, greatly improving their flexibility and adaptability. By combining process control technology with optimized motion control parameters, it achieves dynamic adjustment of speed and path during mowing, thereby improving mowing efficiency and accuracy and effectively avoiding repeated mowing and missed areas.

[0144] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A path planning method for a lawnmower based on process control, characterized in that: include, The system performs an environmental scan of the lawnmower's operating area to obtain point cloud data of the environment and the spatial location of obstacles within the mowing area, generates a real-time environmental map of the mowing area, and performs obstacle detection. Based on the acquired point cloud data and the spatial location of obstacles, an occupation raster map and a lawn state raster map are obtained, and a preliminary path for the mowing area is generated based on the occupation raster map and the lawn state raster map. Based on the initial path and the initial state of the lawnmower, a reference velocity sequence is generated for each waypoint on the initial path, and the motion control parameters of the lawnmower are set. During the lawnmower's execution path, the movement status of the lawnmower and the surrounding environment are monitored in real time, the safe distance between the lawnmower and obstacles is calculated, and path risk indicators are obtained. Based on the path risk indicators, the lawnmower path is replanned. Based on the acquired real-time obstacle information and the lawnmower's motion feedback, the path and motion control parameters of the lawnmower are dynamically adjusted.

2. The path planning method for a lawnmower based on process control as described in claim 1, characterized in that: The process of generating a real-time environmental map of the mowing area includes scanning the environment using LiDAR to obtain point cloud data of the environment where the lawnmower is located. The system acquires environmental images via a camera and processes them using a deep convolutional neural network to determine the spatial location of obstacles. Point cloud data and the spatial location of obstacles are fused using a Kalman filter to generate a real-time environmental map of the mowing area.

3. The path planning method for a lawnmower based on process control as described in claim 2, characterized in that: The obstacle detection refers to calculating the avoidance path in real time by comparing the current path of the lawnmower with the location of obstacles in the real-time environmental map of the mowing area, and using a path planning algorithm.

4. The path planning method for a lawnmower based on process control as described in claim 3, characterized in that: The process of obtaining the occupancy grid map and the lawn state grid map includes projecting point cloud data and the spatial position of obstacles onto a ground coordinate system to obtain a two-dimensional occupancy grid map. The occupied grid is expanded to extend the obstacle boundaries to a safe distance, forming a no-entry zone. The obtained grass height and density are used as lawn state variables and mapped to a raster of the same resolution to obtain a lawn state raster map. Align the lawn status raster map with the occupied raster map according to the raster index.

5. The path planning method for a lawnmower based on process control as described in claim 4, characterized in that: The preliminary path for generating the mowing area includes calculating the required coverage intensity on each passable grid cell and determining the sweep line step size based on the required coverage intensity. Connect the ends of adjacent sweep lines in a reciprocating sequence, and the connecting segment adopts a combination trajectory of circular arc and straight line that satisfies the minimum turning radius constraint; If the connecting section intersects with the restricted area, a detour connecting section is generated along the passable side of the restricted boundary, and the corresponding waypoint is updated; Calculate the distance field based on the occupancy grid map; Perform an iterative smoothing on the waypoint sequence. Without breaking through the restricted area and the safety expansion zone, move the curvature change points away from the obstacles in the opposite direction of the distance field gradient, and ensure that the spacing between adjacent points and the change in direction meet the lawnmower motion constraints. After smoothing, output the preliminary path.

6. The path planning method for a lawnmower based on process control as described in claim 5, characterized in that: Based on the initial path and the initial state of the lawnmower, a reference velocity sequence is generated for each waypoint on the initial path, and the motion control parameters of the lawnmower are set. The specific steps are as follows: The lawnmower obtains its initial state before starting to execute the initial path; The arc length and curvature of each connecting segment of the preliminary path waypoint sequence are calculated sequentially, and the corresponding lawn load is obtained by sampling at the waypoints based on the required cover intensity. Curvature and lawn load constraints are set for the lawnmower speed to generate the upper bound of the reference linear velocity for each waypoint and form a reference velocity sequence; Discretize the upper bound of the reference linear velocity to a sequence of waypoints to obtain the reference velocity sequence; Based on the initial state, calculate the lateral and heading deviations of the lawnmower relative to the initial path starting point, determine the target convergence time based on the magnitude of the lateral and heading deviations, and calculate the lateral control parameters based on the target convergence time. Determine speed control parameters based on a reference speed sequence; The motion control parameters of the lawnmower include lateral control parameters and speed control parameters.

7. The path planning method for a lawnmower based on process control as described in claim 6, characterized in that: The specific steps for calculating the safe distance between the lawnmower and the obstacle, and obtaining the path risk index, are as follows: At the beginning of each control cycle, the lawnmower collects point cloud data, camera images, wheel speed odometer data, and inertial measurement data. By fusing wheel speed odometer and inertial measurement data, the current planar position, heading angle, linear velocity and angular velocity of the lawnmower are obtained, and the current state is used as the starting state for the path update of this cycle. The lateral deviation and heading deviation of the lawnmower relative to the current mowing path are calculated. Ground segmentation is performed on point cloud data, non-ground points are retained and clustered to obtain candidate obstacle outlines; Target detection is performed on the camera images to obtain obstacle bounding boxes, and the obstacle bounding boxes are projected into two-dimensional contours in the ground coordinate system to obtain the obstacle set for the current period; Assign a unique identifier and velocity estimate to each obstacle in the obstacle set for this period; Write the set of obstacles for this cycle into the occupied grid map, calculate the safe distance from any position to the nearest obstacle from the obstacle boundary point, and sample the distance along the forward preview arc length of the current mowing path to obtain the path risk index.

8. The path planning method for a lawnmower based on process control as described in claim 7, characterized in that: The lawn mowing route replanning based on path risk indicators includes... Within the same control cycle, each sampling point within the forward preview arc length is checked. If the distance obtained by substituting the coordinates of a path point into the safety distance field is less than or equal to zero, the current path is determined to be infeasible under the safety margin, and local replanning is immediately triggered. The local replanning includes generating an obstacle bypass connection segment within the passable area of ​​the grid map, starting from the current estimated state and ending at the starting point of the next target coverage segment in the current coverage sequence, and splicing it with the subsequent coverage segment of the original mowing path. After splicing, the connecting segments are smoothed by curvature constraints to obtain the updated waypoint sequence; The updated waypoint sequence, corresponding segment type information, and forward reference speed sequence are written into the tracking reference value of the next control cycle, enabling the lawnmower to update while executing.

9. The path planning method for a lawnmower based on process control as described in claim 8, characterized in that: The specific steps are as follows: Based on the acquired real-time obstacle information and the lawnmower's motion feedback, the path and motion control parameters of the lawnmower are dynamically adjusted. Project the current state onto the current execution path and obtain the lateral deviation, heading deviation, velocity deviation, and curvature following error as motion feedback quantities; The tracking deviations of path risk indicators and motion feedback are normalized on the same scale and then averaged integrally to obtain joint triggering indicators. Within the same control cycle, the path and motion control parameters of the lawnmower are dynamically adjusted based on the triggering rule determination of the joint triggering index.

10. The path planning method for a lawnmower based on process control as described in claim 9, characterized in that: The triggering rules include: if any point within the forward preview range satisfies the condition that the distance obtained by substituting the path point coordinates into the safety distance field is less than or equal to zero, then the current path is determined to be infeasible, and local replanning and parameter tuning are performed. Otherwise, if the joint triggering index is greater than or equal to the replanning threshold, then perform local replanning and parameter tuning; Otherwise, if the combined triggering index is greater than or equal to the parameter tuning threshold, parameter tuning will be performed; Otherwise, keep the mowing path unchanged and only update the reference amount for the next cycle; The parameter adjustment includes continuously reducing the reference speed upper limit and steering gain parameters by a preset scaling factor in each control cycle.

Images