Multi-row photovoltaic panel intelligent cleaning equipment operation path planning method

Abstract

The invention belongs to the field of low-speed intelligent unmanned driving, and discloses an intelligent cleaning path planning method for multiple rows of photovoltaic panels. A heuristic method based on Reeds-Shepp path generation is used to generate a short-distance path, an A * graph search technology is combined to generate a globally optimal long-distance path, and the short-distance path and the long-distance path are integrated according to photovoltaic panel arrangement conditions by referring to intelligent equipment vehicle kinematics model constraints and vehicle cleaning states to generate an overall path. The method and the equipment have extremely high environmental adaptability, can overcome the influence of various actual factors such as the arrangement spacing of the photovoltaic panels, the height difference between the head and the tail of the photovoltaic panels and the complex terrain, and can be flexibly applied to the working environments of various photovoltaic stations. The method is not only efficient, but also has good generalization ability and good robustness.

Description

Technical Field

[0001] The present invention belongs to the field of low-speed intelligent driving, and particularly relates to the field of intelligent cleaning path planning for multi-row photovoltaic panels. Background Art

[0002] Photovoltaic power generation is an environmentally friendly and renewable clean energy, and is an important part of China's new energy strategy. Dust and rainwater residues and other impurities are likely to accumulate on the surface of photovoltaic panels, which will significantly affect the light absorption efficiency and thus reduce the power generation efficiency. In order to make the photovoltaic power station operate efficiently, daily cleaning and maintenance are crucial. Most existing power generation stations use automated equipment for cleaning. In the prior art, generally, an offline tool is used manually to generate a cleaning path map and then sent to the working equipment. For example, by using RTK point positioning and relying on the method of personnel point marking to build a map, the positions of the photovoltaic panels are marked. The equipment operator needs to manually draw a cleaning driving path around the photovoltaic panels according to the arrangement of the photovoltaic panels and based on the measured safety distance, and set cleaning key point marks, start, end and other operation marks.

[0003] However, this method faces the following problems: First, the layouts of different photovoltaic power stations are different, the specifications and sizes of the photovoltaic panels are different, and the distances between rows and the widths are also different. The design and adjustment of the operation path require manual intervention and repeated inspections to ensure that the drawn path is error-free. The flexibility is poor, the operation is complex, and it will consume manpower, material resources and time costs. Second, manual drawing is inaccurate and rough, and it is impossible to accurately ensure the cleaning operation length and safety distance, and there is no way to ensure the safety and effectiveness of the path. Third, in the face of complex scenarios, such as turning in a limited space or turning around in the middle of a narrow corridor, it is very difficult to calculate an optimal curve that meets the motion constraints, and it is difficult to meet the requirements of efficient automated cleaning.

[0004] Therefore, a method that can solve the cleaning path planning problem at one time is needed. Summary of the Invention

[0005] The object of the present invention is to provide an intelligent and efficient method for planning the operation path of a multi-row photovoltaic panel intelligent cleaning device.

[0006] To achieve the above object, the present invention provides the following solution: An intelligent and efficient method for planning the operation path of a multi-row photovoltaic panel intelligent cleaning device, including the following steps:

[0007] Step 1: Select the photovoltaic panels to be cleaned according to the map information of the photovoltaic power station;

[0008] Step 2: Translate 1.5 to 2 meters outward along the bottom edge of each row of photovoltaic panels to generate the cleaning path from the starting point to the ending point of each row of photovoltaic panels; first, sort the coordinate information of each row of photovoltaic panels to be cleaned according to the map information of the photovoltaic power station to obtain the cleaning operation row list of the photovoltaic panels.

[0009] Step 3: Use the heuristic path generation method to connect the ending point of each section of the cleaning path with the starting point of the next row of the cleaning path to generate the non-cleaning path;

[0010] Step 4: Perform a rapid geometric collision detection on each section of the non-cleaning path;

[0011] Step 5: When the number of non-cleaning paths passing the geometric collision detection is 1, generate the corresponding short-distance path; when the number of non-cleaning paths passing the geometric collision detection is 0, use the graph search algorithm to extend the path to the main road and calculate the shortest path required for moving between two points on the main road to generate the corresponding long-distance path;

[0012] Step 6: Connect all the cleaning paths, short-distance paths, and long-distance paths to generate the overall walking path of the automatic intelligent device.

[0013] Sort the coordinate information of each row of photovoltaic panels to be cleaned according to the map information of the photovoltaic power station to obtain the cleaning operation row list of the photovoltaic panels.

[0014] According to the parity of the row index number, calculate the cleaning progress direction. With the front facing the photovoltaic panel, set the cleaning direction of several rows of photovoltaic panels to be from left to right or from right to left, and set the cleaning direction of the even rows of photovoltaic panels to be opposite to that of the odd rows.

[0015] On the bottom edge line of each row of photovoltaic panels, translate 1.5 to 2 meters outward and extend 2 to 4 meters forward and backward for the working distance. Set the cleaning function marks at the starting point and ending point of the path to generate the straight cleaning path from the starting point to the ending point of each row.

[0016] Set cleaning marks for the cleaning path to guide the intelligent device to deploy the left or right mechanical cleaning arm at the starting point of the work and retract the cleaning arm when reaching the ending point of the path until it moves to the starting point of the next cleaning path.

[0017] From the ending point of the cleaning path to the starting point of the next row of the cleaning path, we construct a heuristic path generation method based on the Reeds-Shepp template to generate the shortest path for a vehicle with a turning radius from the starting point to the ending point. This path is composed of a series of straight line segments and arc segments, where:

[0018] Straight line segment : The vehicle drives straight without a change in direction.

[0019] , where is the starting point, and

[0020] is the ending point. The arc segment : The vehicle travels along the arc with the minimum turning radius , divided into left turns , or right turns , and the angle change is

[0021]

[0022] The list of path combinations is shown in Table 1:

[0023] Table 1

[0024] In Table 1, indicates that the vehicle moves forward while turning left; indicates that the vehicle moves backward while turning left; indicates that the vehicle moves forward while turning right; indicates that the vehicle moves backward while turning right; indicates that the vehicle moves straight forward; indicates that the vehicle moves straight backward.

[0025] Considering the cleaning environment of the photovoltaic power station, the short - distance path curves between multiple rows use the following path combinations, called the basic templates:

[0026] Template 1: : , which is a combination of straight line, turn, straight line, turn, and straight line.

[0027] Template 2: :: , which is a combination of three arc segments.

[0028] To simplify the calculation, we transform the whole process through coordinate transformation into , and the formed path. The calculated polar coordinate diagram is shown in Table 2:

[0029] Table 2

[0030] In Table 2: represents the relative abscissa difference represents the relative ordinate difference represents the relative radian difference represents the first left / right turn angle ​Indicates the straight-line reverse distance Indicates the angle of the second turn in the same direction Calculated as:

[0031]

[0032]

[0033] The path is: straight line (a), turn (t), straight line (u), turn (v), straight line (b).

[0034] Using the above formula, between the end point of the current row of photovoltaic panels and the starting point of the next row, we substitute multiple sets of straight-line distances a and b to find possible solutions.

[0035] During the path planning process, to ensure that there is no collision between each path point and the photovoltaic panel and to meet the constraint requirements of the intelligent vehicle's geometric shape, we propose a fast calculation method as follows:

[0036] Step 1: Construct a spatial partitioning BSP structure for obstacles to accelerate the detection speed.

[0037] Step 2: For each path point, generate the geometric contour of the intelligent vehicle.

[0038] Step 3: If the geometric contour of the intelligent vehicle is outside the working area, then eliminate this path.

[0039] Step 4: Calculate whether the geometric contour of the intelligent vehicle is within the expanded bounding box of the photovoltaic panel. If not, continue to detect the next point.

[0040] Step 5: Use the GJK method for polygon collision detection to determine whether there is a collision between the photovoltaic panel and the geometric contour of the intelligent vehicle. If there is a collision, eliminate this path.

[0041] To ensure the accurate lateral working distance between the intelligent device and the photovoltaic panel, before the cleaning path operation, a straight section of 2 to 4 meters in length is introduced in the planning to allow the intelligent device to adjust its posture to align with the next straight-line operation.

[0042] During this process, the intelligent device retreats a certain distance to adjust its posture. The longitudinal distance is x and the lateral distance is y.

[0043] This process can be expressed by the formula: rotation (ϕ) → right turn (t) → straight line (u) → reverse turn (t).

[0044] That is, rotation in different directions: .

[0045] The specific formula is as follows: Let x be the lateral distance from the current position to the straight-ahead position, y be the perpendicular distance from the current position to the target line, ϕ be the horizontal deviation angle, and t be the rotation angle.

[0046] It can be seen that when the path is: rotate (a), turn right (t), go straight (u), reverse (t).

[0047]

[0048]

[0049]

[0050] When there is no solution for the short-distance path, a long-distance path will be generated.

[0051] Starting from the starting point and the ending point of the cleaning path, extend them to the main road respectively, search for the search nodes on the main road that are closest to the starting point and the ending point of the cleaning path, and thus use the graph search A* algorithm to search for the shortest path between these two points on the main road network. The connections of these paths can be smoothed with a curve of the rotation radius to form a long-distance path.

[0052] First, starting from the starting point, search nodes are evenly distributed on the main road to find the nodes on the nearest main road. The search nodes closest to the starting point and the ending point of the cleaning path are found using distance calculation. If a search node is blocked by an obstacle and does not meet the collision detection, it is connected to other reachable nodes.

[0053] Then, use the A* algorithm to search for the shortest path between the main road nodes.

[0054] The main road network can be represented as a weighted directed graph .

[0055] Among them, represents the set of all search nodes on the main road, E represents the set of connecting edges between nodes, and the search cost estimation function is , where: is the actual cost to reach the current node x, is the estimated cost from node x to the target node, that is, the Euclidean distance.

[0056] The steps to find the long-distance path are as follows:

[0057] Step 1: Create a queue Q for dynamically selecting the current node with the minimum cost for expansion. The cost of all nodes is zero and they are all in the open state.

[0058] Step 2: Obtain the current node x with the minimum cost from the queue Q. If this node is the starting node, the loop ends.

[0059] Step 3: Close node x, write the cost value, and remove it from the queue. Select the non-closed node with the minimum estimated cost of the adjacent edges of this node and add it to the queue. Repeat: Step 2.

[0060] Step 4: Backtrack the path: Starting from the target node, select the node with the minimum cost and backtrack to the starting node to form a sequence of path nodes.

[0061] Step 5: Connect between the path nodes with a curve to generate a smooth path that complies with the composite kinematic rules.

[0062] Finally, connect all the cleaning paths, short-distance paths, and long-distance paths to generate the overall walking path of the automatic intelligent device.

[0063] The present invention has the following beneficial effects:

[0064] First, it realizes the efficient cleaning path planning of intelligent devices in different photovoltaic panel row spacing scenarios.

[0065] Second, it is not only applicable to standard photovoltaic power stations, but also applicable to complex scenarios with obstacles nearby, generating intelligent walking paths, reducing manual intervention, and improving operation efficiency. Description of the Drawings

[0066] Figure 1 is the flowchart of the path generation of the present invention;

[0067] Figure 2 is the manually planned operation path in the background art;

[0068] Figure 3 is the map information of a single block of the photovoltaic power station yard;

[0069] Figure 4 is the straight cleaning path generated by the present invention;

[0070] Figure 5 is the path generation template of the present invention;

[0071] Figure 6 is the movement schematic diagram of the intelligent device of the present invention during the process of aligning with the straight line;

[0072] Figure 7 is the schematic diagram of the non-cleaning path extending to the main road of the present invention;

[0073] Figure 8 is the result of the photovoltaic cleaning path planning generated by the present invention.

[0074] Description of reference numerals: T1 and T3 are the photovoltaic panels to be cleaned, T2 is a non-cleaning photovoltaic panel, and B1 is an obstacle. R1 represents the main road. The circles n1 - n7 on the road are the graph search nodes on the road, and the remaining multiple paths p1 - p7 are the short paths connecting the cleaning points to the main road. Detailed implementation manners

[0075] Obtain the map information of a single block of the photovoltaic power station through a GPS positioning system, RTK navigation data, or a ground survey tool, and extract the coordinate information of the photovoltaic panels to be cleaned from within the operation block of the photovoltaic power station, such as Figure 3 As shown, the rows of horizontal black long strips are the photovoltaic panels, arranged in rows facing south. The sets of unfilled convex polygons represent irregular obstacles. The outer border represents the area of the entire block, and the main road runs along the border.

[0076] Sort the photovoltaic panels in the same row in the horizontal direction, and then sort them in the north-south direction to ensure that each row of photovoltaic panels is arranged in the order from south to north.

[0077] Then, generate a straight-line operation list for the bottom edges of the photovoltaic panels based on the coordinate information of each row of photovoltaic panels. This list can calculate the starting point and ending point of the bottom edge of each row based on the coordinate data of the photovoltaic panels.

[0078] Set the cleaning direction of travel according to the parity of the row index. For odd rows, clean from left to right; for even rows, clean from right to left. The cleaning route adopts a bow-shaped route. During the travel process, the device will translate 1.5 meters to 2 meters outward in a straight line for the cleaning distance and extend 2 meters to 4 meters forward and backward to ensure that the entire surface of the photovoltaic panel is covered.

[0079] Set cleaning marks for the cleaning path to guide the intelligent device to deploy the left or right mechanical cleaning arm at the starting point of the work and retract the cleaning arm when reaching the end point of the path until it moves to the starting point of the next cleaning path.

[0080] Through the above steps, generate the cleaning path for each row. As Figure 4 shown, the cleaning sequence is row by row. There may be several rows of photovoltaic panels in each row, and the cleaning directions of each row are opposite. The cleaning path is a straight line slightly longer than the photovoltaic panel.

[0081] Generation of Short Path: From the end point of the cleaning path to the starting point of the cleaning path for the next row, we construct a heuristic path generation method based on the Reeds-Shepp template, which can be called the short path. The Reeds-Shepp path generation was proposed by Jack Reeds and Lawrence Shepp in 1990 and is used to describe the shortest path of a vehicle with a turning radius from the starting point to the end point. This path is composed of a series of straight line segments and circular arc segments and can be expressed as follows:

[0082] 1. Straight line segment : The vehicle travels straight without a change in direction.

[0083] , where is the starting point, is the end point.

[0084] 2. Circular arc segment : The vehicle travels along a circular arc with the minimum turning radius , divided into a left turn or a right turn , and the angle change is .

[0085]

[0086] The path combination list is shown in Table 1:

[0087] Table 1

[0088] Indicates that the vehicle moves forward with a left turn; Indicates that the vehicle moves backward with a left turn; Indicates that the vehicle moves forward with a right turn; Indicates that the vehicle moves backward with a right turn; Indicates that the vehicle moves forward straight; Indicates that the vehicle moves backward straight.

[0089] Considering the cleaning environment of the photovoltaic power station, the following path combinations are used for the short path curves between multiple rows, which are called the basic templates:

[0090] 1. : , a combination of straight line, turn, straight line, turn, and straight line.

[0091] 2. : , a combination of three circular arcs.

[0092] To simplify the calculation, we transform the whole process through coordinate transformation into , the formed path is the same as the above method. The calculated polar coordinate diagram is shown in Table 2:

[0093] Table 2 In Table 2: represents the relative abscissa difference represents the relative ordinate difference represents the relative radian difference represents the first left / right turn angle represents the straight-line backward distance represents the second same-direction turn angle

[0094] Calculated:

[0095]

[0096]

[0097] Using the above formula, between the end point of the current row of photovoltaic panels and the starting point of the next row, we substitute multiple sets of straight-line distances a and b to find possible solutions.

[0098] As Figure 5 shown, there are 5 cases for the short-path generation template from the previous row to the next row of photovoltaic panels. Figure 5 Among them, the arrows on the path indicate the walking direction of the intelligent device. T1 and T3 are the photovoltaic panels that need to be cleaned, T2 is the non-cleaning photovoltaic panel, and B1 is the obstacle. (a) is the path of turning left and going straight; (b) is the short-path of turning right and going straight and then turning right; (c) is the path of operating on every other row of photovoltaic panels; (d) is the short-path of turning left and going straight and then turning left; (e) is the operation path planned to bypass the obstacle, the case of CSCSCSC.

[0099] Geometric constraint detection:

[0100] During the path planning process, to ensure that there is no collision between each path point and the photovoltaic panel and to meet the constraint requirements of the geometric shape of the intelligent vehicle, we propose a fast calculation method as follows:

[0101] Step 1: Construct a spatial partitioning BSP structure for obstacles to accelerate the detection speed.

[0102] Step 2: For each path point, generate the geometric contour of the intelligent vehicle.

[0103] Step 3: Whether the geometric contour of the intelligent vehicle is outside the operation area. If so, eliminate this path.

[0104] Step 4: Calculate whether the geometric contour of the intelligent vehicle is within the expanded photovoltaic panel bounding box. If not, continue to detect the next point.

[0105] Step 5: Use the GJK method for polygon collision detection to determine whether there is a collision between the photovoltaic panel and the geometric contour of the intelligent vehicle. If there is a collision, eliminate this path.

[0106] To ensure the accurate lateral working distance between the intelligent device and the photovoltaic panel, a straight section with a length of 2 to 4 meters is introduced in the path planning before the cleaning path operation, enabling the intelligent device to adjust its posture to align with the subsequent straight-line operation.

[0107] As Figure 6 shown, during this process, the intelligent device retreats a certain distance to adjust its posture, with the longitudinal distance being x and the lateral distance being y.

[0108] This process can be expressed as a formula: Rotate (ϕ) → Turn right (t) → Straight line (u) → Reverse (t).

[0109] That is, rotation in different directions: .

[0110] The specific formula is as follows:

[0111] Let x be the lateral distance from the current position to the straightening position, y be the vertical distance from the current position to the target straight line, ϕ be the horizontal deviation angle, and t be the rotation angle.

[0112] It can be known that , when , the path is: Rotate (a), Turn right (t), Straight line (u), Reverse (t).

[0113]

[0114]

[0115]

[0116] Generation of long-distance paths:

[0117] When there is no solution for the short-distance path, generate a long-distance path. Starting from the starting point and the ending point of the cleaning path, extend them to the main road respectively, search for the nodes on the main road closest to the starting point and the ending point of the cleaning path, and then use the A* algorithm for graph search to find the shortest path between these two points on the main network. Smooth these paths into a curve with a rotation radius and connect them to form a complete long-distance path.

[0118] As Figure 7As shown in the figure, R1 represents the main road. The circles n1 - n7 on the road are the graph search nodes on the road, and the remaining multiple paths p1 - p7 are the short - distance paths connecting the cleaning points to the main road.

[0119] First, starting from the starting point, search nodes are evenly distributed on the main road to find the nearest node on the main road. Using distance calculation, find the search nodes closest to the starting and ending points of the cleaning path. If the search node is blocked by an obstacle and does not meet the collision detection, connect to other reachable nodes.

[0120] Then, use the A* algorithm to search for the shortest path between the main road nodes.

[0121] The main road network can be represented as a weighted directed graph , represents the set of all search nodes on the main road, E represents the set of connecting edges between nodes, and the search cost estimation function is , where: is the actual cost to reach the current node x, is the estimated cost from node x to the target node, that is, the Euclidean distance.

[0122] The algorithm steps are as follows:

[0123] Step 1: Create a queue Q for dynamically selecting the current node with the minimum cost for expansion. The cost of all nodes is zero and they are all in the open state.

[0124] Step 2: Obtain the current node x with the minimum cost from the queue Q. If this node is the starting node, the loop ends.

[0125] Step 3: Close node x, update the cost value, and remove it from the queue. Select the non - closed node with the minimum cost estimation of the adjacent edges of this node and add it to the queue, and repeat: Step 2.

[0126] Step 4: Backtrack the path: Starting from the target node, select the node with the minimum cost and backtrack to the starting node to form a sequence of path nodes.

[0127] Step 5: Use curves to connect between the path nodes to generate a smooth path that complies with the compound kinematic rules.

[0128] Finally, connect all the cleaning paths, short - distance paths, and long - distance paths to generate the overall walking path of the automatic intelligent device, as Figure 8 shown.

Claims

1. An intelligent planning method for operating paths of unmanned intelligent equipment, characterized in that: The following steps are involved: Step 1: According to the map information of the photovoltaic station, select the photovoltaic panels that need to be cleaned; Step 2: Move horizontally 1.5 to 2 meters outward along the bottom edge of each row of photovoltaic panels to generate a cleaning path from the starting point to the end point of each row of photovoltaic panels; Step 3: Use the heuristic path generation method to connect the end point of each cleaning path with the starting point of the next row of cleaning paths to generate a non-cleaning path; Step 4: Perform fast geometric collision detection on each non-sweeping path; Step 5: When the number of non-sweeping paths that pass the geometric collision detection is 1, the corresponding short-distance path is generated; when the number of non-sweeping paths that pass the geometric collision detection is 0, the path is extended to the trunk road using the graph search algorithm, and the shortest path required to move between two points on the trunk road is calculated to generate the corresponding long-distance path; Step 6: Connect all cleaning paths, short-distance paths and long-distance paths to generate the overall walking path of the automatic intelligent device.

2. The short-distance path according to claim 1, characterized in that: It is a motion path heuristically generated based on the Reeds-Shepp motion model, which contains an effective combination of straight movement and turning.

3. The long distance route according to claim 1, characterized in that: A graph search algorithm was used to find the shortest path connecting the end point of one row of sweeping paths to the starting point of the next row on the main roads within the site, and the path was smoothed using a rotation radius curve.

4. The geometric collision rapid detection technology according to claim 1 is characterized in that: First, use the global BSP search tree, and then use the bounding box collision detection algorithm to check whether there is a collision between the shape of the smart device and the expansion distance of the photovoltaic panel and other obstacles. After eliminating the path that will cause geometric collision, use the polygon collision detection GJK algorithm to calculate whether the shape of the smart device will collide with the photovoltaic panel and other obstacles.

Images