Cross-domain cluster unmanned collaborative lotus root picking and collecting method and system

Abstract

The invention discloses a cross-domain cluster unmanned collaborative lotus root picking and collecting method and system. According to the method, a map is sequentially executed in blocks, that is, global path planning is converted into local path planning, and a main task is decomposed into a plurality of sub-tasks; in each sub-task, firstly, position information of lotus roots and obstacles is determined through an existing detection and recognition method, then, based on an artificial potential field method, a gravitational field is set for the positions of the lotus roots, a repulsive force field is set for the positions of the obstacles, and then the lotus roots are gradually moved to the positions of the lotus roots by bypassing the obstacles in the total potential field gradient descending direction to be dug; and the purpose of detecting and identifying at the same time is achieved. The sub-tasks are executed in sequence, and the lotus root digging task is completed. According to the method, the whole lotus root field can be traversed comprehensively, the path can be adjusted adaptively according to real-time sensor data and environment changes, so that collision with obstacles is avoided, smooth completion of tasks is ensured, the coverage is large, and the robustness is good.

Description

Technical Field

[0001] The present invention relates to the technical fields of lotus root picking, path planning, and agricultural mechanization, and particularly relates to a cross-domain cluster unmanned collaborative lotus root picking and collection method and system. Background Art

[0002] In the field of lotus root picking technology, the traditional manual method of digging lotus roots relies on the skills and experience of workers to determine the best picking sequence and position of lotus roots. This method is inefficient and labor-intensive. With the development of unmanned aerial vehicle (UAV) technology, a global path planning method based on UAV mapping has been proposed. This method uses sensors such as sonar and radar carried by UAVs to scan and map the lotus root field, obtaining map information of the entire lotus root field, including the positions of lotus roots and obstacles. Then, using the map path planning algorithm, an optimal path is formulated for digging lotus roots based on the information of the positions of lotus roots and obstacles. However, since lotus roots are buried in the soil and often covered by lotus leaves, and the turbidity of the water body also affects the measurement accuracy of the sensors, it is difficult to obtain complete information on the distribution of lotus root joints and obstacles, resulting in an incomplete global map construction with low accuracy. At the same time, existing path planning methods usually take the shortest path as the optimization goal and have insufficient adaptability in actual complex environments. Summary of the Invention

[0003] In view of this, the present invention provides a cross-domain cluster unmanned collaborative lotus root picking and collection method and system, which realizes planning based on detection through existing underwater sensing technology, thereby upgrading the traditional global static planning to a global-local dynamic planning system; and based on a multi-objective optimization model, a global trend guiding line is generated by comprehensively considering parameters such as path efficiency, mechanical energy consumption, and obstacle avoidance safety factor, which can effectively improve the automation level and efficiency of lotus root picking.

[0004] The cross-domain cluster unmanned collaborative lotus root picking and collection method of the present invention includes:

[0005] Step 1: Using a UAV to collect the coordinate set of the boundary points of the lotus root field, and then combining with the detection radius of the sensors of the lotus root digging machine itself to generate a probability grid map;

[0006] Overlay an initial Gaussian noise on the probability grid map, and then based on multiple sets of data collected by the UAV, combine with the least squares method to calculate the error and update the value of the Gaussian noise to obtain an accurate map;

[0007] Perform main route planning on the accurate map using a fuzzy map path planning algorithm; wherein, the cost function of the fuzzy map path planning algorithm is:

[0008] F(n) = α·G(n) + β·H(n) + γ·D(n)

[0009] Among them, G(n) represents the actual distance from the starting point to node n, H(n) represents the heuristic distance from node n to the end point, and D(n) represents the fuzzy membership degree of node n to the obstacle; α, β, and γ are weights, which are adjusted using the Q-learning algorithm;

[0010] Step 2: Divide the main route at intervals of the detection radius of the self-sensor of the lotus root digging machine to obtain each pause point;

[0011] Step 3: The lotus root digging machine moves forward along the main path to the pause position; during the forward movement, the detection of lotus roots and obstacles is carried out synchronously; at the same time, a dynamic safety redundancy distance is set. If the distance between the lotus root digging machine and the obstacle or lotus root is less than or equal to the set dynamic safety redundancy distance, the local path planning is used to bypass the obstacle or lotus root;

[0012] Step 4: At the pause position, set the positions of the detected lotus roots as the gravitational field, and the positions of the obstacles as the repulsive field, and superimpose them to construct the total potential field map of the local environment and calculate its potential field gradient; the lotus root digging machine gradually moves to the positions of each lotus root along the direction of the potential field gradient descent to dig and collect the lotus roots; after completing the digging and collection of the last lotus root in the local environment, return to the pause position along the original path;

[0013] Step 5: The lotus root digging machine continues to move forward along the main path, repeating Step 3 to Step 5 until it reaches the end point.

[0014] Preferably, in Step 1, the initial Gaussian noise is taken as 0.3.

[0015] Preferably, in Step 3, the dynamic safety redundancy distance is:

[0016] d safe =R s +0.2σ

[0017] Wherein, R s is the detection radius of the self-sensor of the lotus root digging machine; σ is the variance of the Gaussian noise.

[0018] Preferably, in Step 3, when an obstacle or lotus root is detected within the dynamic safety redundancy distance, the artificial potential field method, the VFH algorithm or the Dijkstra algorithm is used for local path planning to bypass the obstacle or lotus root.

[0019] Preferably, in Step 3, the lotus root digging machine detects the lotus roots and obstacles underwater and on the water surface respectively; then the underwater and water surface detection data are fused to obtain the unified three-dimensional positions of the lotus roots and obstacles.

[0020] Preferably, a sonar sensor, an ultrasonic sensor or a capacitive sensor is used for underwater detection; a lidar, a multispectral imaging sensor or a thermal imaging sensor is used for water surface detection.

[0021] Preferably, a collection vehicle is also provided; when the lotus root digger is full of lotus roots, the collection vehicle plans a path to the lotus root digger by using the artificial potential field method according to the current position of the lotus root digger and the obstacle position determined by the lotus root digger, and transfers the lotus roots collected by the lotus root digger to a designated location.

[0022] The present invention also provides a cross-domain cluster unmanned collaborative lotus root picking and collection system, including a lotus root digger, wherein a sensor and a path planning unit are provided on the lotus root digger; the sensor is used to collect the position information of the lotus roots and obstacles on and under the water surface; the path planning unit plans a path by using the above method to complete the lotus root picking and collection.

[0023] Beneficial effects:

[0024] The present invention performs sequential block execution on the map, that is, transforms the global path planning into local path planning and decomposes the main task into several sub-tasks; in each sub-task, first, the position information of the lotus roots and obstacles is determined by using the existing detection and recognition method, and then based on the artificial potential field method, a gravitational field is set for the position of the lotus roots and a repulsive field is set for the obstacle position, and then move step by step around the obstacle along the gradient descent direction of the total potential field to the position of the lotus roots to dig the lotus roots, achieving the purpose of detecting and recognizing while digging. Each sub-task is executed sequentially to complete the lotus root digging task. The present invention can comprehensively traverse the entire lotus root field, and can adaptively adjust the path according to the real-time sensor data and environmental changes, so as to avoid collisions with obstacles, ensure the smooth completion of the task, and has a large coverage area and good robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a flow chart of the lotus root picking and collection method of the present invention.

[0026] Figure 2 It is a main route planning diagram.

[0027] Figure 3 It is a flow chart for implementing the artificial potential field algorithm of the present invention.

[0028] Figure 4 It is a total potential field map.

[0029] Figure 5 It is a successful lotus root digging path using the traditional method.

[0030] Figure 6 It is the planned path of the present invention.

[0031] Figure 7 It is a lotus root digging failure path due to obstacles using the traditional method. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0032] The following examples are given in conjunction with the drawings to describe the present invention in detail.

[0033] The present invention provides a cross - domain cluster unmanned collaborative lotus root picking and collection method. As Figure 1 shown, first, using the range of the sensors of the lotus root digging machine itself and the map information of the lotus root field boundary obtained by the UAV scanning, a main route is autonomously planned. This main route is the overall path for the lotus root digging machine to pick in the lotus root field. Through this main route, the lotus root digging machine can efficiently cover the entire lotus root field, reducing ineffective movement and waiting time. Then, several pause points are planned on the main route. The lotus root digging machine performs local scanning at the pause points to obtain accurate local map information of the lotus roots and obstacles, and based on the artificial potential field method, local multi - target path planning is carried out to obtain a local lotus root digging path. After completing the lotus root digging in the local map, it returns to the main route and goes to the next planned pause point to perform the same work until the last planned pause point is completed. When the pressure sensor detects that the collection box of the lotus root digging machine is full during the lotus root digging process, the collection vehicle will start from the starting point and use the same artificial potential field method path planning as the lotus root digging vehicle, scanning the environment while planning until it reaches the lotus root digging machine. Thus, the cross - domain collaborative method of the lotus root digging machine, collection vehicle and UAV is realized, and efficient lotus root digging is completed.

[0034] Specifically, it includes the following steps:

[0035] Step 1, plan the main route.

[0036] First, use the UAV to collect the coordinate set {P i (x i , y i )} of the lotus root field boundary points, and then based on the detection radius R s of the sensors of the lotus root digging machine itself, combine the boundary points to generate a polygon area and generate a probability grid map.

[0037] Then, Gaussian noise is superimposed on the probability grid map to simulate map uncertainty. Specifically, first set an initial value, σ = 0.3m, and then use the data obtained from multiple UAV scans to train the noise model parameter σ, use the least - squares method to fit the model data and calculate the error, and dynamically update the σ value to improve the map accuracy. Finally, an accurate map as Figure 2 shown is drawn.

[0038] Next, an improved fuzzy map path planning algorithm is used to plan the main route. Define the cost function:

[0039] F(n)=α·G(n)+β·H(n)+γ·D(n)

[0040] Among them, G(n) represents the actual distance from the starting point to node n, H(n) represents the heuristic distance from node n to the end point, and D(n) represents the fuzzy membership degree of node n to the obstacle. α, β, and γ are weights, and the present invention uses the Q-learning algorithm to adjust them, with the goal of weighted path smoothness and energy consumption.

[0041] Step 2: Determine the pause positions planned on the main route.

[0042] Taking the detection range of the sensor as the interval, divide the main route to obtain each pause position, as Figure 2 shown.

[0043] Step 3: The lotus root harvester moves forward along the main path to the pause position; during the forward movement, the lotus root harvester synchronously detects obstacles and lotus roots on the water surface and underwater; at the same time, a dynamic safety redundancy distance d safe = R s + 0.2σ is set. If the distance between the lotus root harvester and the obstacle or lotus root is less than or equal to the set dynamic safety redundancy distance, local path planning is performed using methods such as the artificial potential field method, VFH algorithm, Dijkstra algorithm, etc. to bypass the obstacle or lotus root.

[0044] As Figure 2 shown, after determining the main path and the starting point and the end point, the positive direction and the negative direction can be obtained. The "positive direction" is the moving direction along the main path. While moving forward, underwater and water surface detections are carried out to determine the positions of lotus roots and surrounding obstacles. Among them, for complex underwater / muddy conditions, ultrasonic sensors such as sonar can be used, which can penetrate sediment and effectively identify the shapes and distributions of lotus roots / obstacles, and effectively detect and identify the positions of lotus roots and obstacles.

[0045] In this embodiment, an integrated underwater and water surface detection technology combining sonar and lidar is adopted to realize the detection and position construction of lotus roots and obstacles within a certain range. In the underwater / muddy part, the lotus root harvester is equipped with a sonar system. By emitting ultrasonic waves and receiving the echoes reflected from underwater obstacles or lotus root stalks, the distance information of the target is obtained. According to the time difference of ultrasonic wave propagation, the depth and position of the underwater target can be calculated, thereby constructing a two-dimensional or three-dimensional map of the possible positions of underwater lotus root stalks, obstacles, etc., covering a certain detection range. In the water surface part, the lotus root harvester uses lidar technology. By emitting laser beams in the same direction and receiving the signals reflected by the laser, the spatial positions of water surface lotus root stalks, lotus leaves, and obstacles are obtained. A map of the water surface environment is generated. Finally, the map data of the water surface and underwater are fused. Through data registration and fusion, the errors of different detection methods are eliminated, and a unified map of the positions of lotus root stalks and obstacles is constructed.

[0046] Step 4, at the pause position, based on the detected local lotus root targets and the position information of obstacles, the artificial potential field method is used to perform local multi-target point path planning and complete lotus root digging.

[0047] After obtaining the positions of local lotus root targets and obstacles, the artificial potential field method (APF) is used for path planning. The core idea is to model the environment where the lotus root digger is located as a potential field. Among them, the target point, that is, the lotus root, generates an attractive force on the lotus root digger, while the obstacle generates a repulsive force on the lotus root digger. Through the action of these two forces, the lotus root digger calculates the potential energy gradient and then automatically plans a path to avoid obstacles and reach the target point, as Figure 3 shown.

[0048] Specifically, as shown in Equation (1), a gravitational field is set for the position of the lotus root, and as shown in Equations (2) and (3), a repulsive field is set for the position of the obstacle. The shape and intensity of the gravitational field and the repulsive field will change with the distance between the lotus root digger and the target point and the obstacle. The closer to the target, the smaller the attractive force, and the closer to the obstacle, the greater the repulsive force.

[0049]

[0050] V 斥力 = 0, ρ (lotus root digger, obstacle) > ρ0 (3)

[0051] Among them, ρ is the distance between the lotus root digger and the lotus root target / obstacle, and ρ0 is the safe distance between the lotus root digger and the obstacle, which is set as a fixed value according to the actual size of the lotus root digger.

[0052] After setting the gravitational field and the repulsive field for each lotus root target position and each obstacle position, all the gravitational field and repulsive field vectors are superimposed to obtain the total potential field map as Figure 4 shown.

[0053] Based on the artificial potential field method, the gradient descent direction of the total potential field represents the direction of the resultant force on the lotus root digger at the current position, that is, the path direction of moving from the high potential energy region to the low potential energy region. The lotus root digger starts from the current pause point and moves step by step along the gradient descent direction of the total potential field ( Figure 4 ), and the speed depends on the magnitude of the gradient. The step size is dynamically adjusted by the PID control algorithm to ensure smooth movement. The repulsive field makes the lotus root digger automatically deviate from the obstacle area, and the path shows a detour trajectory. The lotus root digger reaches each lotus root target position in turn and completes lotus root digging.

[0054] After the lotus root digger completes lotus root digging at the last lotus root target position in the local map, it returns to the pause position of the main route along the original path.

[0055] Step 5, the lotus root digger continues to move forward along the main path to the next pause position, and repeats steps 3 to 5 until it reaches the end point.

[0056] To improve the efficiency of lotus root digging, a collection vehicle can be used to transfer the lotus roots collected by the lotus root digger to the collection area, eliminating the need for the digger to travel to the collection area after filling up and then return to continue digging. When the digger is full of lotus roots, a collection vehicle is directly dispatched to collect the lotus roots. The collection vehicle uses the artificial potential field method for path planning, which is the same as in Step 4.

[0057] In this invention, E-puck robots are used for simulation in Webots. The E-puck robots adopt a design with front wheels and rear tracks, and a simulation and comparative analysis is carried out between the method of this invention and the traditional manual lotus root digging method and the lotus root digging method based on global map path planning by drone scanning. Figure 5 It is the path to successfully dig lotus roots using traditional manual experience; Figure 6 It is the path planned by the method of this invention. As for the lotus root digging method based on global map path planning by drone scanning, according to on-site research, it is found that due to the turbid water, the shielding of lotus leaves, and the interference of soil media, the drone cannot directly obtain the global map of lotus roots and obstacles. Therefore, this invention believes that in the process of lotus root digging, the method of using drones for one-time global detection and positioning is not feasible, so no simulation analysis is carried out on this method.

[0058] The simulation results are shown in Table 1. Table 2 is the comparison of output predictions in actual situations.

[0059] Table 1 Comparison table of output in the simulation environment

[0060] Traditional method (successful) Global map path planning Fuzzy map path planning Feasibility Feasible Infeasible Feasible Time efficiency 2”11’ 8”54’ Coverage rate 75％ 100％ Robustness Poor Good Energy consumption Less More Path length Short Long

[0061] Table 2 Comparison table of output predictions in actual situations

[0062] Traditional method Global map path planning Fuzzy map path planning Feasibility Feasible Infeasible Feasible Time efficiency Cannot be determined Can be determined Coverage rate 50％-80％ 100％ Robustness Poor Good Energy consumption Cannot be determined Can be determined Path length Cannot be determined Can be determined

[0063] As can be seen from Table 1 and Table 2, the traditional method may have advantages in terms of time efficiency, energy consumption, and path length, but these data are obtained under the condition of no encounter with obstacles and smooth lotus root digging according to the specified route. However, in practical applications, the traditional method often faces various obstacles or other unknown situations, resulting in the inability to continue digging lotus roots along the original route, such as Figure 7As shown, this may cause the lotus root digger to need to return to re-dig the lotus roots, change the digging direction, or even collide with obstacles and get damaged. These problems will significantly increase the time required for lotus root digging and reduce the quantity of dug lotus roots. In contrast, the method of the present invention shows obvious advantages in these aspects. By utilizing the fuzzy map, the lotus root digger can comprehensively traverse the entire lotus root field and plan the path according to the map information, thus achieving a higher coverage rate. In this way, the lotus root digger can intelligently plan the path and flexibly adjust according to the specific situation of the lotus root field to ensure maximum coverage of every part of the lotus root field. In addition, the method of the present invention can adaptively adjust the path according to real-time sensor data and environmental changes, thereby avoiding collisions with obstacles and having better robustness. This is very important for preventing unexpected situations of the lotus root digger and ensuring the smooth completion of the task.

[0064] In summary, the above are only the preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A cross-domain cluster unmanned collaborative lotus root picking and collection method, characterized in that, Including: Step 1: Use a drone to collect the coordinate set of the lotus root field boundary points, and then generate a probability grid map in combination with the detection radius of the sensors of the lotus root digging machine itself. Overlay an initial Gaussian noise on the probability grid map, and then based on the multiple acquisition data of the drone, combine the least squares method to calculate the error and update the value of the Gaussian noise to obtain an accurate map. Use the fuzzy map path planning algorithm to perform the main route planning on the accurate map; among them, the cost function of the fuzzy map path planning algorithm is: F(n) = α·G(n) + β·H(n) + γ·D(n) where G(n) represents the actual distance from the starting point to node n, H(n) represents the heuristic distance from node n to the end point, D(n) represents the fuzzy membership degree of node n to the obstacle; α, β, γ are weights and are adjusted using the Q-learning algorithm. Step 2: Divide the main route at intervals of the detection radius of the sensors of the lotus root digging machine itself to obtain each pause point. Step 3: The lotus root digging machine moves forward along the main path to the pause position; during the forward movement of the lotus root digging machine, the detection of lotus roots and obstacles is carried out synchronously; at the same time, a dynamic safety redundancy distance is set. If the distance between the lotus root digging machine and the obstacle or lotus root is less than or equal to the set dynamic safety redundancy distance, the artificial potential field method is used for local path planning to bypass the obstacle or lotus root. Step 4: At the pause position, set the positions of the detected lotus roots as the gravitational field, and the positions of the obstacles as the repulsive field, and perform superposition to construct a local environment total potential field map and calculate its potential field gradient; the lotus root digging machine gradually moves to the positions of each lotus root along the direction of the potential field gradient descent for digging and collection; after completing the digging and collection of the last lotus root in the local environment, return to the pause position along the original route. Step 5: The lotus root digging machine continues to move forward along the main path, repeating steps 3 to 5 until it reaches the end point.

2. The method according to claim 1, wherein In step 1, the initial Gaussian noise is taken as 0.

3.

3. The method according to claim 1, wherein In step 3, the dynamic safety redundancy distance is: d safe = R s + 0.2σ where R s is the detection radius of the sensors of the lotus root digging machine; σ is the variance of the Gaussian noise.

4. The method according to claim 1 or 3, characterized in that, In step 3, when an obstacle or lotus root is detected within the dynamic safety redundancy distance, the artificial potential field method, the VFH algorithm or the Dijkstra algorithm is used for local path planning to bypass the obstacle or lotus root.

5. The method according to claim 1, characterized in that, In step 3, the lotus root digging machine respectively detects the lotus roots and obstacles under the water and on the water surface; then the detection data under the water and on the water surface are fused to obtain the unified three-dimensional positions of the lotus roots and obstacles.

6. The method according to claim 5, wherein Use a sonar sensor, an ultrasonic sensor or a capacitance sensor for underwater detection; use a lidar, a multi-spectral imaging sensor or a thermal imaging sensor for water surface detection.

7. The method according to claim 1, wherein There is also a collection vehicle; when the lotus root digging machine is full of lotus roots, the collection vehicle plans a path to the lotus root digging machine using the artificial potential field method according to the current position of the lotus root digging machine and the obstacle positions determined by the lotus root digging machine, and transports the lotus roots collected by the lotus root digging machine to a designated location.

8. A cross-domain cluster unmanned collaborative lotus root picking and collection system, characterized in that, Including a lotus root digging machine, on which there are sensors and a path planning unit; the sensors are used to collect the position information of the lotus roots and obstacles on the water surface and under the water; the path planning unit plans a path using the method according to any one of claims 1 to 7 to complete the lotus root picking and collection.

Images