Search

CN-122018510-A - Dynamic path planning and obstacle avoidance method for greenhouse robot

CN122018510ACN 122018510 ACN122018510 ACN 122018510ACN-122018510-A

Abstract

The invention provides a dynamic path planning and obstacle avoidance method for a greenhouse robot, and belongs to the technical field of agricultural robots. The method comprises the steps of collecting multi-source heterogeneous data, carrying out data synchronization and preprocessing, generating a reference track through a path planning algorithm based on the preprocessed multi-source heterogeneous data, predicting a controller through a self-adaptive model based on the reference track, setting constraint conditions by taking the minimum cost function as a target and using physical limitation and safety requirements of a robot, and carrying out iteration through an optimization algorithm to output an optimal control sequence. According to the invention, through on-line model parameter identification, the AMPC can adapt to the dynamic change and environmental disturbance of the robot, and more accurate track tracking control is provided compared with a fixed model controller, and particularly, the effect is remarkable when the load is changed or the ground is slipped.

Inventors

  • SHANG HANG
  • LI ZIXUAN
  • ZHANG YANSONG
  • LI XIANG
  • SHEN RENJIE

Assignees

  • 大连工业大学艺术与信息工程学院

Dates

Publication Date
20260512
Application Date
20260305

Claims (6)

  1. 1. A dynamic path planning and obstacle avoidance method for a greenhouse robot is characterized by comprising the following steps: Acquiring multi-source heterogeneous data, and carrying out data synchronization and preprocessing; And predicting a controller through a self-adaptive model based on the reference track, setting constraint conditions by taking the minimum cost function as a target and setting physical limit and safety requirements of the robot, iterating by utilizing an optimization algorithm, and outputting an optimal control sequence.
  2. 2. The method of claim 1, wherein generating the reference trajectory by a path planning algorithm based on the preprocessed multi-source heterogeneous data comprises: By A The algorithm plans a collision-free initial path connecting the starting point and the target point through the global path planner, and the collision-free initial path is expressed as a path sequence: Wherein, the Comprising , Is that The desired orientation of the points; B spline fitting is carried out on the discrete path point sequence, and a continuous reference track taking arc length s as a parameter is obtained: Wherein s is a path arc length parameter; the X coordinate of a point with the arc length s on the track under the global coordinate system is represented, namely the horizontal position of the robot at the point; The Y coordinate of a point with the arc length s on the track under the global coordinate system is represented, namely the vertical position of the robot at the point; The tangential angle of the trajectory at the point of arc length s is indicated, at which point the robot should have an ideal orientation.
  3. 3. The method of claim 1, wherein the cost function: Wherein, the The predicted state of the robot at the moment k+i; the reference state is the reference state at the corresponding moment; Q and R are given positive weight matrixes, and are used for balancing tracking precision and control smoothness; is the distance between the predicted robot and the jth obstacle; for a preset safe distance threshold value, Punishing weights for obstacles; In order to predict the length of the time domain, To control the time domain length, M is the number of obstacles.
  4. 4. The method of claim 1, wherein the constraint comprises: Wherein, the Representing the future state X-direction coordinates of the robot, Representing the future state X-direction coordinates of the dynamic barrier, Representing the Y-direction coordinates of the future state of the robot, Representing the future state Y-direction coordinates of the dynamic barrier.
  5. 5. The method of claim 1, wherein the multi-source heterogeneous data comprises: greenhouse environment data and robot state data; The greenhouse environment data comprises crop row spacing, soil firmness, temperature and humidity and illumination intensity; the robot state data includes pose, speed, wheel speed and load.
  6. 6. The method of claim 1, wherein the preprocessing comprises: Filtering and denoising the laser radar point cloud, fusing camera data and segmenting a scene, and extracting elements such as ground, crop rows and barriers; the multispectral image identifies passable areas and barriers through correction and semantic segmentation; And the inertial measurement unit and the encoder data are fused through Kalman filtering to generate smooth real-time pose and speed estimation of the robot.

Description

Dynamic path planning and obstacle avoidance method for greenhouse robot Technical Field The invention relates to the technical field of agricultural robots, in particular to a dynamic path planning and obstacle avoidance method for greenhouse robots. Background In modern agricultural greenhouses, robots are core equipment for performing operations such as seeding, inspection, spraying and the like. The advantages and disadvantages of the path planning and obstacle avoidance capability directly relate to the operation efficiency and the crop safety. The current greenhouse robots mostly adopt a preset fixed path mode, and an environment map is constructed by collecting two-dimensional coordinates of crop furrows and static facility position information. The method cannot effectively cope with dynamic changes in the greenhouse, such as temporary entering operators, movable irrigation equipment or line spacing changes caused by crop growth, manual intervention is often required to adjust paths, the efficiency is low, and collision or crop rolling is easy to occur. In addition, the existing obstacle avoidance system depends on a single sensor, is incomplete in perception, has poor flexibility among dense crop rows, needs to reserve a larger safety distance, and reduces the space utilization rate. Meanwhile, the control of the system on the cooperation of the obstacle avoidance and the operation is simple, the obstacle avoidance is usually triggered only by a distance threshold value, and the differences of the crop growth state, the ambient light, the ground flatness and the like cannot be considered, so that the operation is easy to interrupt, the crop is damaged or the equipment is easy to fail in a complex scene. And the existing control method has limited processing capacity for system constraints (such as robot dynamics constraint and environment physical constraint), and is difficult to realize real-time optimization in a dynamic environment on the premise of ensuring safety. Model Predictive Control (MPC) can deal with constraint and optimization problems, but the model of the traditional MPC is fixed, so that the model is difficult to adapt to the dynamic parameter changes (such as wheel slip and load change) of robots in greenhouse environments and the severe changes of environmental characteristics. Therefore, a greenhouse robot dynamic path planning and obstacle avoidance method which can cope with dynamic environments, achieve accurate obstacle avoidance, guarantee operation coordination and have environment self-adaptation capability is needed. Disclosure of Invention In view of the above, the invention provides a dynamic path planning and obstacle avoidance method for a greenhouse robot, which acquires greenhouse environment data through a multidimensional sensor, builds a nonlinear prediction model by combining the state of the robot to obtain a prediction state, and determines an objective function through the prediction state to perform optimization iteration to perform path planning. For this purpose, the invention provides the following technical scheme: A dynamic path planning and obstacle avoidance method for greenhouse robots comprises the following steps: Acquiring multi-source heterogeneous data, and carrying out data synchronization and preprocessing; And predicting a controller through a self-adaptive model based on the reference track, setting constraint conditions by taking the minimum cost function as a target and setting physical limit and safety requirements of the robot, iterating by utilizing an optimization algorithm, and outputting an optimal control sequence. Further, the generating the reference track based on the preprocessed multi-source heterogeneous data through a path planning algorithm comprises the following steps: By A The algorithm plans a collision-free initial path connecting the starting point and the target point through the global path planner, and the collision-free initial path is expressed as a path sequence: Wherein, the Comprising,Is thatThe desired orientation of the points; B spline fitting is carried out on the discrete path point sequence, and a continuous reference track taking arc length s as a parameter is obtained: Wherein s is a path arc length parameter; the X coordinate of a point with the arc length s on the track under the global coordinate system is represented, namely the horizontal position of the robot at the point; The Y coordinate of a point with the arc length s on the track under the global coordinate system is represented, namely the vertical position of the robot at the point; The tangential angle of the trajectory at the point of arc length s is indicated, at which point the robot should have an ideal orientation. Further, the cost function: Wherein, the The predicted state of the robot at the moment k+i; the reference state is the reference state at the corresponding moment; Q and R are given positive weight matrixes, and are