Navigation system for unmanned forklifts in conditions with poor positioning

Abstract

The present invention discloses an unmanned forklift navigation system suitable for conditions with poor positioning effect. The construction method is as follows: importing the location of cargo and the size of cargo column; using micro grids to divide the two-dimensional ground plane, requiring each grid to record its own risk value cost risk ; For the cost in the grid map risk The small grid with a value of 1 is puffed in the sense of computer graphics, and the cost of the small grid occupied by the puffing operation is risk The values are all set to 1; for the remaining costs in the grid map risk The positioning effectiveness of each tiny grid cell, each with a value of 0, is then verified. The grid map is then constructed into a connected graph with a standard eight-connected structure. The A* search algorithm is then used to traverse this connected graph, starting from the starting node in the connected graph and continuing until the end node is found. This invention achieves accurate and reliable autonomous forklift navigation under conditions where indoor infrared positioning effectiveness is limited, without the need for denser deployment of infrared sensors.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned forklifts, and in particular to an unmanned forklift navigation system suitable for use under conditions where positioning effects are poor. Background Art

[0002] Indoor logistics and warehousing environments are important application scenarios for autonomous driving. To achieve automation in indoor logistics and warehousing, forklifts that transport goods are often operated in an unmanned manner around the clock, that is, unmanned forklifts, also known as autonomous driving forklifts.

[0003] The unmanned forklift system, like other unmanned vehicle systems, has on-board modules including positioning, navigation, perception, decision-making and control. The positioning function is to perceive the vehicle's own position information based on sensors, which directly provides support and guarantee for the navigation, decision-making, control and other modules of the unmanned vehicle.

[0004] Classic positioning methods for indoor scenarios include WiFi positioning, infrared positioning, ZigBee positioning, and radio frequency identification positioning. Infrared solutions are often used in logistics warehouses or workshops, providing accurate location information for autonomous vehicles operating indoors. However, infrared positioning technology has strict requirements for indoor lighting conditions and obstructions. If these requirements are not met, positioning results will be unsatisfactory. For example, infrared positioning technology requires a relatively clear indoor environment; otherwise, the infrared beam is easily blocked by indoor obstacles. Severe obstruction can completely render the positioning method ineffective. As can be seen from the above, positioning is a critical function in autonomous driving systems, and its effectiveness must be guaranteed.

[0005] To improve the reliability of indoor infrared positioning, existing technology involves densely deploying infrared sensors at various high points indoors to ensure that a sufficient number of infrared sensors emit unobstructed infrared light that reaches the vehicle at all times. However, the problem with existing technology is that densely deploying infrared sensors significantly increases the cost of implementing positioning functions.

[0006] In order to prevent positioning from causing an economic burden, this patent will propose a new technology, namely, to achieve accurate and reliable forklift autonomous driving navigation function under limited positioning capabilities without the need for more densely deployed infrared sensor equipment. Summary of the Invention

[0007] The technical problem to be solved by the present invention is that, in order to solve the problem that the indoor infrared positioning effect is not ideal under occlusion conditions, resulting in poor performance of the automatic driving system, an unmanned forklift navigation system suitable for conditions with poor positioning effect is proposed.

[0008] This is a navigation system for unmanned forklifts that is suitable for conditions with poor positioning. The construction method includes:

[0009] Step 1: Import the cargo placement and cargo column dimensions in the warehouse environment;

[0010] Step 2: For the two-dimensional ground plane, use a resolution scale of ΔX hc ×ΔY hc It is divided into small grids and then a grid map is formed, requiring each grid to record its own risk value cost risk , and the cost of each grid risk The value is initialized to 0;

[0011] Step 3: The cost of the grid occupied by the goods risk Set to 1;

[0012] Step 4: Cost in the grid map risk The small grid with a value of 1 is expanded in the sense of computer graphics, and the expansion radius is set to R, and the small grid cost occupied by the expansion operation is risk The value is also always set to 1;

[0013] Step 5: For the remaining costs in the grid map risk Take the tiny grids with a value of 0 and test their positioning effects one by one;

[0014] Step 6: Construct the grid map into a connected graph with a standard eight-connected structure, and then call the A* search algorithm to traverse the connected graph from the starting node in the connected graph until the end node is found.

[0015] In a preferred embodiment of this scheme, the cargo placement position and cargo column size in step one are uniformly maintained and collected through a background data system. When the cargo enters the warehouse, its length, width and height are recorded. After the cargo is accurately placed in the warehouse by a forklift, the three-dimensional space occupied by the cargo can be accurately known.

[0016] In a preferred embodiment of this solution, the expansion radius R in step 4 is the radius length of the area occupied by the forklift on the ground covered by the circular envelope of the forklift's exterior.

[0017] In a preferred embodiment of this solution, the method for checking the positioning effect one by one in step 5 is:

[0018] Assuming that the coordinates of the geometric center point of a small grid on the ground plane are (m,n), the cost recorded by the small grid is risk The value should be determined as follows:

[0019]

[0020] The above formula ensures that the cost recorded in each small grid risk The maximum value does not exceed 1, because it is the minimum value obtained by adding a certain operation value to 1; this value decreases as the number of devices that can be illuminated by the small grid center point (m,n) increases. Its theoretical minimum value is (4-N beam ).

[0021] In a preferred embodiment of this solution, once the A* search algorithm traverses the cost risk The small grids with a value of 1 are considered illegal nodes and closed, and the remaining small grids are considered passable nodes. The A* search algorithm is specifically:

[0022] f=g+w as cost risk ;

[0023] Where f represents the complete cost of the node, g represents the cumulative cost from the starting point to the node, and cost risk What is recorded is the risk value of the current node, w as >0 is the weight parameter specified by the user; cost risk The value is between (4-N beam ) and 0, and the smaller the value is, the more infrared transceiver devices can capture the node, so the location perception effect will be better.

[0024] The implementation of the embodiments of the present invention has the following beneficial effects:

[0025] The present invention discloses an unmanned forklift navigation system suitable for conditions with poor positioning effects. Under conditions where indoor infrared positioning effects are limited, accurate and reliable forklift automatic driving navigation functions can be achieved without the need for more densely deployed infrared sensor equipment. The use of this unmanned forklift navigation system can ensure that the forklift always travels in an area with good positioning effects, thereby ensuring that downstream modules such as the vehicle's decision-making and control are not adversely affected, thereby improving the reliability of the automatic driving vehicle system. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0027] Figure 1 This is a top view of the forklift;

[0028] Figure 2 This is a schematic diagram of the indoor positioning principle of a forklift;

[0029] Figure 3 This is a top view of the indoor warehouse; DETAILED DESCRIPTION

[0030] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0031] See Figure 1 , Figure 1 This is a top view of a forklift; the kinematics of a forklift chassis are similar to those of cars on the market, with the front wheels steering and the rear wheels fixed in direction. Figure 1 In the equation, (x, y) represents the coordinates of the midpoint of the forklift's rear wheel axle, v represents the forklift's speed, φ represents the forklift's front wheel deflection angle, and θ represents the vehicle's attitude angle in the inertial coordinate system. In addition, Figure 1 The vehicle geometric parameters are also indicated, L W is the vehicle wheelbase parameter, L F Characterizes the vehicle's front overhang parameter, L R Characterizes the length of the vehicle's rear suspension, L B represents the vehicle width parameter.

[0032] The area occupied by the forklift on the ground is often covered by the forklift's outer circular envelope, e.g. Figure 1 As shown by the dotted line in the figure, assuming that the radius of this circle is R, since the forklift has a certain height, the forklift as a whole can be regarded as a column moving horizontally on the ground. During the movement, the column cannot collide with obstacles in the environment. H A "reflective marker" was deployed at the location to reflect infrared light.

[0033] For the indoor positioning principle of forklifts, please refer to Figure 2In a warehouse filled with goods at varying heights, several (not many) infrared transceivers are deployed on the roof. Each device independently emits a modulated infrared light source. This light is reflected by a reflective marker on the top of a forklift and recaptured by the device. By measuring the time difference between the reflected light and the incident light, the distance between the reflective marker and the device can be estimated to be twice the distance. Theoretically, if the reflective marker can be detected simultaneously by three devices, the coordinates of the reflective marker on the top of the forklift can be uniquely determined in three dimensions, based on the principle of three-point positioning (this principle states that if the distances of any point P in space to three known points A, B, and C are known, then point P is determined and uniquely located), thus completing the positioning task. Taking into account the potential measurement errors and complex real-world factors such as light dissipation and interference in three-dimensional space, it is generally required that the reflective marker be illuminated by at least four devices simultaneously. Locations that meet this condition are considered ideal for positioning, while those that do not are considered poorly positioned. We expect vehicles to operate in ideal positioning locations.

[0034] Figure 2 Several cylindrical cargoes of different heights are placed in the middle room, and a total of 8 devices are deployed at the commanding heights of the room. When the vehicle is at a certain position, the line connecting the reflective marker on its top and the various devices may be blocked by taller cargo. Because light travels in a straight line, the device cannot illuminate the reflective marker under obstruction conditions. In the vehicle position shown in the figure, the reflective marker on the roof can be illuminated by 5 devices at the same time, and the remaining 3 devices cannot be illuminated because they are blocked by cargo. The conclusion is that the vehicle can be accurately located when it is in this position. The following is a mathematical expression of whether the vehicle can be accurately located:

[0035] Assume that N are deployed in the indoor area. beam There are three infrared transceiver devices, and the infrared light capture point of the kth device is located in the three-dimensional space. Coordinates, these coordinate information are known in advance. Let us assume that the reflective marker on the roof is located at coordinates (x w ,y w ,L H ), then the necessary and sufficient condition for the accurate positioning of the point is that the point can be illuminated by at least 4 devices. This condition can be expressed as:

[0036]

[0037] Among them, the function IsReachable() is used to determine whether the distance between two coordinate points in three-dimensional space is blocked by obstacles; the parameter format of the function IsReachable() is IsReachable([a,b,c],[d,e,f]), which is used to determine whether the two coordinate points M=(a,b,c) and N=(d,e,f) in three-dimensional space are blocked by obstacles in the environment. If there is an obstruction, the function IsReachable=0, otherwise the function takes the value of 1. The implementation idea of the function IsReachable() is to discretize the line segment MN into N according to a certain resolution. sample The algorithm selects sufficiently dense and equally spaced sampling points, and then checks each of these sampling points to see if they fall within the obstacles in the three-dimensional environment space. To determine whether a sample point Q in a space falls within an environmental obstacle, it only needs to check whether point Q falls within the cargo envelope.

[0038] See Figure 3 , Figure 3 This is a bird's-eye view of an indoor warehouse. Given a starting point and a destination, the navigation system is responsible for determining a rough route that connects the starting and ending points continuously. The figure shows several alternative routes that effectively connect the starting and ending points. A key difference between these routes is that they belong to different homotopy classes. This is a concept in topological geometry: two lines in space that can migrate toward each other without crossing obstacles (cargo) belong to the same homotopy class. As shown in the figure, the four lines are separated by obstacles and therefore belong to four different homotopy classes. Some homotopy classes correspond to shorter routes, allowing the vehicle to reach the destination earlier, thus saving efficiency; conversely, some homotopy classes correspond to longer routes. Furthermore, some routes corresponding to these homotopy classes inevitably pass through areas with inaccurate positioning, which we want to avoid. Considering route length (efficiency), positioning accuracy, and collision safety, the navigation system should provide a route corresponding to the optimal homotopy class to connect the starting and ending points.

[0039] Therefore, this application proposes an unmanned forklift navigation system suitable for conditions with poor positioning effect, and the construction method includes:

[0040] Step 1: Import the cargo placement and column dimensions in the warehouse environment. These are centrally maintained and collected through a backend data system. When cargo enters the warehouse, its length, width, and height are recorded. Once the cargo is accurately placed in the warehouse by a forklift, the three-dimensional space occupied by the cargo can be accurately determined.

[0041] Step 2: For the two-dimensional ground plane, use a resolution scale of ΔX hc ×ΔYhc It is divided into small grids and then a grid map is formed, requiring each grid to record its own risk value cost risk , and the cost of each grid risk The value is initialized to 0. It should be noted that the indoor area is a three-dimensional area, but we will focus on the two-dimensional ground plane for the time being (note that the ground is two-dimensional).

[0042] Step 3: The cost of the grid occupied by the goods risk Set to 1;

[0043] Step 4: Cost in the grid map risk The small grid with a value of 1 is expanded in the sense of computer graphics, and the expansion radius is set to R (i.e. Figure 1 The circle radius R shown by the dotted line in the middle) and the small mesh cost occupied by the expansion operation risk The value is always set to 1.

[0044] The purpose of performing the puffing operation in step 4 is because the forklift is not a point mass on a two-dimensional plane, but is instead considered a disk with a radius of R. Therefore, if the puffing process is not performed, it is necessary to consider that the vehicle disk does not collide with obstacles in the environment. If the vehicle disk is shrunk toward the center of the circle with a shrinkage scale of R, and at the same time the obstacles in the environment are puffed (the expansion scale is also R), then this expansion and contraction change is equivalent. At this point, the vehicle can be regarded as a point mass. If the point mass does not collide in the puffed environment, then the vehicle body will also not collide in the original environment.

[0045] Step 5: For the remaining costs in the grid map risk Take the tiny grids with a value of 0 and test their positioning effect one by one. The specific test method is:

[0046] Assuming that the coordinates of the geometric center point of a small grid on the ground plane are (m,n), the cost recorded by the small grid is risk The value should be determined as follows:

[0047]

[0048] The above formula ensures that the cost recorded in each small grid risk The maximum value does not exceed 1, because it is the minimum value obtained by adding a certain operation value to 1; this value decreases as the number of devices that can be illuminated by the small grid center point (m,n) increases. Its theoretical minimum value is (4-N beam ).

[0049] Step 6: Construct the grid map into a connected graph with a standard eight-connected structure, and then call the A* search algorithm to traverse the connected graph from the starting node in the connected graph until the end node is found.

[0050] Once the A* search algorithm traverses the cost risk The small grids with a value of 1 are considered illegal nodes and closed, and the remaining small grids are considered passable nodes. risk Nodes with a value of 1 are considered illegal because the vehicle risks collision or loss of positioning capability if it reaches these points. The A* search algorithm produces a broken line connecting the start and end points. This line avoids obstacles, bypasses areas with poor positioning, and has a shorter path length under the same conditions.

[0051] When using the A* search algorithm, we have made targeted improvements to it so that it can generate the optimal homology category. When calculating the cost value of the nodes traversed in the connected graph, the proposed improved A* search algorithm no longer uses the original f=g+h calculation scheme of the A* search algorithm, but changes it to

[0052] f=g+w as cost risk ;

[0053] Where f represents the complete cost of the node, g represents the cumulative cost from the starting point to the node, and cost risk What is recorded is the risk value of the current node, w as >0 is a weight parameter specified by the user. Compared with the classic f=g+h solution, the Manhattan distance formula is no longer used to calculate the heuristic estimate h from the current node to the end point, thus ensuring that the cost estimate of each node is conservative, thus ensuring that high-quality results are obtained in the sense of the shortest path length. Based on removing the heuristic estimate h, this solution further increases the cost risk For valid nodes, cost risk The value is between (4-N beam ) is a non-positive number between 0 and 0. The smaller the value, the more infrared transceiver devices can capture the node, so the location perception effect will be better. risk The purpose of the price is to encourage vehicles to choose areas where they can obtain location awareness signals more smoothly under the same conditions.

[0054] Based on statistical data from long-term use, the use of this system has increased the accuracy of positioning of unmanned forklifts during operations in warehouses by 192%, a significant improvement.

[0055] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A navigation system for unmanned forklifts suitable for use under conditions of poor positioning, characterized in that: Its construction method includes: Step 1: Import the cargo placement and cargo column dimensions in the warehouse environment; Step 2: For the two-dimensional ground plane, use a resolution scale of It is divided into small grids and then a grid map is formed, requiring each grid to record its own risk value , and each grid's The value is initialized to 0; Step 3: The grid occupied by the goods Set to 1; Step 4: Target the grid map The small grid with a value of 1 is expanded in the sense of computer graphics, and the expansion radius is set to R, and the small grid occupied by the expansion operation is expanded. The value is also always set to 1; Step 5: For the remaining ones in the grid map Take the tiny grids with a value of 0 and test their positioning effects one by one; The method to test the positioning effect one by one is: Assume that the coordinates of the geometric center point of a small grid on the ground plane are , the height from the top of the forklift to the ground is , let the The coordinates of the infrared light capture point of the road equipment in three-dimensional space are , then the microgrid records The value should be determined as follows: ； The above formula ensures that each tiny grid records The maximum value does not exceed 1, because it is the result of a certain operation value and 1 to get the min operation; this value changes with the center point of the tiny grid. The theoretical minimum value is ; Step 6: Construct the grid map into a connected graph with a standard eight-connected structure, and then call the A* search algorithm to traverse the connected graph from the starting node in the connected graph until the end node is found; The A* search algorithm is specifically: ； in represents the complete cost value of the node, represents the cumulative cost from the starting point to the node, and What is recorded is the risk value of the current node. is the weight parameter specified by the user; The value is between A non-positive number between 0 and 0. The smaller the value, the more infrared transceiver devices can capture the node, so the location perception effect will be better.

2. The unmanned forklift navigation system suitable for use under conditions of poor positioning according to claim 1, characterized in that: The placement of the goods and the dimensions of the cargo columns in step 1 are uniformly maintained and collected through a background data system. When the goods enter the warehouse, their length, width, and height are recorded. After the goods are accurately placed in the warehouse by a forklift, the three-dimensional space occupied by the goods can be accurately known.

3. The unmanned forklift navigation system suitable for use under conditions of poor positioning according to claim 1, characterized in that: The expansion radius R in step 4 is the radius of the area occupied by the forklift on the ground covered by the circular envelope of the forklift's exterior.

4. The unmanned forklift navigation system suitable for use under conditions of poor positioning according to claim 1, characterized in that: In step 6, once the A* search algorithm traverses The tiny grids with a value of 1 are considered illegal nodes and are closed, and the remaining tiny grids are considered passable nodes.

Images