Bridge crane safety anti-collision system and method based on laser radar

Abstract

The invention relates to the technical field of automatic control of cranes, and discloses a bridge crane safety anti-collision system and method based on laser radar. Comprising the following steps: step 1, acquiring point cloud data when a bridge crane bridge works through a laser radar, and preprocessing the point cloud data; 2, environment object segmentation is carried out; and step 3, establishing an anti-collision control strategy. According to the method, self-positioning and environment mapping of the bridge crane are completed by using a laser radar SLAM technology, and the problems that environment detection of a single sensor is low in efficiency, easy to be influenced by illumination, insufficient in detection capability and the like are solved; according to the method, the radar point cloud after environment mapping is subjected to clustering segmentation, so that collision detection is facilitated; according to the method, the anti-collision model based on the collision distance is established, the anti-collision disposal strategy is designed according to the anti-collision model, and anti-collision disposal can be completed hierarchically; the problems that a traditional bridge crane is insufficient in environment detection capacity and load hoisting collision protection, and cannot complete self position positioning and surrounding environment map building are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of automatic control of cranes, and particularly relates to a safety anti-collision system and method for a gantry crane based on lidar. Background Art

[0002] A gantry crane is a heavy-duty material handling equipment widely used in places such as factories, warehouses, and ports. It mainly consists of core components such as a bridge, a trolley traveling mechanism, a hoisting trolley, and a control system. When in use, the gantry crane realizes material handling in three-dimensional space by coordinating the movements of the trolley, the hoisting trolley, and the lifting mechanism. Due to its advantages such as high efficiency, high load-bearing capacity, and good scalability, it is widely used in the industrial production field.

[0003] In the design process of gantry cranes, the current design of the safety anti-collision system for gantry cranes mainly focuses on preventing collisions between bridge frames, and there is less research on collision problems during load hoisting. However, there are still the following key problems in practical applications:

[0004] 1. Insufficient protection against load hoisting collisions. The existing system lacks the ability to detect dynamic obstacles (such as mobile equipment, personnel, and temporary structures) during load hoisting in real time. Especially in complex industrial scenarios (such as steel plants and ports), load swing or path deviation is likely to cause collision accidents. The current anti-collision logic is mostly based on a preset path or a static obstacle database, and it is difficult to adapt to the dynamically changing environment, resulting in frequent emergency braking or false triggering.

[0005] 2. The methods for environmental detection using sensors in the current load anti-collision system for gantry cranes include infrared detection, camera vision, etc. The infrared detection method can only measure the straight-line distance of a certain point in a single direction and cannot truly achieve the detection effect of the transfer site. The camera vision detection has deficiencies such as a relatively short detection distance, inability to work properly under low light conditions, and a viewing angle limitation.

[0006] 3. The existing anti-collision system for gantry cranes can only detect the distance between the load and the obstacle and then execute anti-collision control, and cannot complete its own position positioning and the construction of the surrounding environment map.

[0007] Based on this, the present invention proposes a safety anti-collision system and method for a gantry crane based on lidar to solve the problems existing in the above-mentioned prior art. Summary of the Invention

[0008] Technical Problems to be Solved

[0009] In view of the deficiencies of the prior art, the present invention provides a safety anti-collision system and method for a gantry crane based on lidar to solve problems such as insufficient environmental detection ability, insufficient protection against load hoisting collisions, and inability to complete self-positioning and surrounding environment map construction existing in traditional gantry cranes.

[0010] Technical solution

[0011] To achieve the above object, the present invention provides the following technical solutions:

[0012] In the first aspect, the present invention provides a safety anti-collision method for a gantry crane based on lidar, including:

[0013] Step 1: Obtain point cloud data when the gantry crane bridge is working through lidar, and process the point cloud data using the SLAM algorithm;

[0014] Step 2: After completing the mapping in the loading environment in Step 1, perform environmental object segmentation

[0015] Step 2.1: Use a segmentation algorithm based on RANSAC to perform clustering segmentation on the point cloud;

[0016] Step 2.2: Use the method of enclosing with a rectangular frame to extract obstacle information;

[0017] Step 3: Establish an anti-collision control strategy

[0018] The anti-collision control judgment threshold based on the collision distance is:

[0019]

[0020] where v r is the relative velocity between the suspended load and the obstacle, α0 is the braking acceleration, t1 is the system response time, t2 is the braking force increase stage, d0 is the reserved buffer distance, and D is the distance between the load and the obstacle in front at time t.

[0021] In a preferred embodiment, the process of processing the point cloud data using the SLAM algorithm in Step 1 includes:

[0022] Step 1.1: Preprocess the point cloud data, filter and downsample the lidar point cloud;

[0023] Step 1.2: Perform coordinate system conversion to convert the point cloud information into a space coordinate system with the center of the suspended load as the origin;

[0024] Step 1.3: After completing the coordinate conversion, use the ICP algorithm for point cloud matching and positioning;

[0025] Step 1.4: After using the ICP algorithm for point cloud matching and positioning in Step 1.3, detect whether there is a loop by matching the point clouds between the current frame and the historical frame. After updating the pose estimation, use the optimized pose and point cloud data to construct a three-dimensional map.

[0026] In a preferred embodiment, the process of preprocessing the point cloud data in Step 1.1 includes:

[0027] Step 1.1.1: Input the point cloud information P = {p i | i = 1, 2,..., n}, find the minimum vertex (x min , y min , z min ) and the maximum vertex (x max , y max , z max ). Use these two points as the vertices of the bounding box, establish a coordinate system with the minimum point as the origin, and enclose all the input point clouds.

[0028] Step 1.1.2: Divide the voxel cube, set the side length of the unit cube l = (r, r, r), and calculate the number of voxel cube divisions through the following formula:

[0029]

[0030] where length is the length of the unit cube, width is the width of the unit cube, height is the height of the unit cube, and x max , y max , z max、 x min , y min , z min represent the maximum or minimum coordinate values on the x, y, and z axes respectively;

[0031] Step 1.1.3: Calculate the geometric center of each voxel cube, and replace all the points in the cube with the coordinates of the center point M(x a , y a , z a ). The center calculation formula is:

[0032]

[0033] where x centre , y centre , z centre represent the geometric center coordinate values on the x, y, and z axes respectively; x i , y i , z i represent the coordinate values of each point on the x, y, and z axes;

[0034] Step 1.1.4: Replace all the point clouds within the voxel cube with the geometric center point M(x a , y a , z a ) of the cube, and integrate all the geometric centers within the cubes into new point cloud information.

[0035] In a preferred embodiment, the process of using the ICP algorithm for point cloud matching and positioning in Step 1.3 includes:

[0036] Step 1.3.1: Read two frames of point cloud data obtained by the lidar, select two sets of point cloud data X and Y, and randomly initialize the transformation matrices R and t;

[0037] Step 1.3.2: Find corresponding points. For each point in the point cloud data X, use the distance metric to find the nearest point in B to A as the corresponding point of A;

[0038] Step 1.3.3: Optimize the rotation matrix R and the translation matrix t

[0039] After obtaining the corresponding points, estimate the rotation matrix R and the translation matrix t using the corresponding points; specifically, use the least squares method to estimate and calculate the optimal matrices, and determine the minimization objective function E(R, t) as:

[0040]

[0041] Step 1.3.4: Iteratively calculate the transformation matrices R and t

[0042] When the new transformation matrices R and t are optimized and the positions of some point clouds change, and the positions of their nearest points also change accordingly; return to the method of finding corresponding points in Step 1.2. After obtaining the new corresponding points, go to Step 1.3 to update the rotation matrix R and the translation matrix t again; Steps 1.2 and 1.3 are iterated until the designed iteration termination condition is met and the objective function is less than the convergence error.

[0043] In a preferred embodiment, the process of using the RANSAC-based segmentation algorithm to perform clustering segmentation on the point cloud in Step 2.1 includes:

[0044] Step 2.1.1: First, randomly select three point clouds from all the point cloud data, judge the geometric relationship of the three point clouds. If the three points are collinear and not in the same plane, reselect three point clouds until they are non-collinear and in the same plane, and then calculate the relevant model parameters of the plane formed by them;

[0045] Step 2.1.2: Traverse and calculate the distances from other point clouds to the plane model constructed in Step 2.1.1, compare them with the set distance threshold, and store the information of the point clouds whose distances to the plane model are less than the distance threshold.

[0046] Step 2.1.3: Repeat Step 2.1.1 and Step 2.1.2 until the set maximum number of iterations is reached. Select the model with the most aggregated points as the fitted plane result according to the information stored in Step 2.1.2 to complete plane fitting.

[0047] In a preferred embodiment, Step 2.2 uses the method of surrounding with a rectangular frame, and the process of extracting obstacle information includes:

[0048] Step 2.2.1: First, find the point clouds with the maximum and minimum coordinate values in the x, y, and z axes of the point cloud space scanned by the lidar in the point cloud cluster respectively. Calculate the length, width, and height dimensions of the rectangular bounding box of the obstacle by subtracting the maximum value from the minimum value, and construct the smallest rectangular bounding box that can envelope all the point clouds from the same object:

[0049] length = |x max - x min |

[0050] width = |y max - y min |;

[0051] height = |z max - z min |

[0052] Step 2.2.2: Obtain the centroid position of the obstacle

[0053] Calculate the centroid coordinate point (x, y, z) of the target obstacle by calculating the average values of all point cloud coordinates on the x, y, and z axes:

[0054]

[0055] Step 2.2.3: Obtain the distance information of the obstacle

[0056] Calculate the obstacle distance information D according to the distance from the centroid position (x, y, z) of the rectangular frame to the lidar and the dimensions of the rectangular frame:

[0057]

[0058] In a second aspect, the present invention provides a safety anti-collision system for a gantry crane based on lidar. The safety anti-collision system is a gantry crane anti-collision system, and the gantry crane anti-collision system is used to implement the safety anti-collision method for a gantry crane based on lidar SLAM.

[0059] In a preferred embodiment, the overhead crane safety anti-collision system based on lidar SLAM includes:

[0060] An environment detection module, which is used to detect the loading environment during use and send the obtained point cloud map, real-time positioning, and obstacle information to the anti-collision decision-making module;

[0061] An anti-collision decision-making module, which is used to perform environmental object segmentation on the received point cloud map, real-time positioning, and obstacle information, and formulate an anti-collision decision based on the target detection result and the status information result;

[0062] A bridge crane control module, which is used to receive the anti-collision decision of the anti-collision decision-making module, convert the anti-collision decision into a control command, and perform vehicle speed control and braking control.

[0063] In a preferred embodiment, the anti-collision decision-making module includes:

[0064] A sensor group, which is used to obtain the overhead crane status information and send the overhead crane status information to the industrial control computer; it includes a rope length measurement sensor and a swing angle measurement sensor. The rope length sensor is installed at the tail of the hoisting motor and is used to measure the change in the wire rope length in real time; the swing angle measurement sensor is used to measure the pull rope displacement in real time and convert it into a swing angle;

[0065] An industrial control computer, which is used to receive the overhead crane status information and the target detection result, analyze the working status of the crane and the movement status of the load according to the received sensor information, and send a frequency conversion control instruction to the bridge crane control module.

[0066] In a preferred embodiment, the bridge crane control module is an electrical control system. The electrical control system takes a programmable logic controller PLC as the core and controls the movement direction and speed of the bridge crane based on the drive mode of PLC-inverter-motor.

[0067] Advantages

[0068] Compared with the prior art, the present invention provides an overhead crane safety anti-collision system and method based on lidar, which has the following advantages:

[0069] 1. The present invention uses lidar SLAM technology to complete the self-positioning and environment mapping of the overhead crane, overcoming the problems of low efficiency of single-sensor environment detection, susceptibility to light influence, and insufficient environment detection ability of traditional overhead cranes.

[0070] 2. The present invention clusters and segments the radar point cloud after environment mapping, and designs a rectangular bounding box to frame the obstacles, thus facilitating collision detection and determining a traversable area for the load movement.

[0071] 3. The present invention utilizes a set of anti-collision models based on collision distance and designs anti-collision disposal strategies accordingly, which can complete anti-collision disposal hierarchically, avoid the wear of the bridge crane caused by single braking operation, reduce the inertial effect of emergency braking, and is more auxiliary; it solves the problems existing in traditional bridge cranes, such as insufficient environmental detection ability, insufficient collision protection for load hoisting, inability to complete self-positioning and the construction of the surrounding environment map, etc. Description of the Drawings

[0072] Figure 1 It is a flowchart of the safety anti-collision method for a bridge crane based on lidar SLAM of the present invention;

[0073] Figure 2 It is an axial diagram of the coordinates of the center point of the anti-collision system and the center point of the lifted object after the installation of the anti-collision system of the present invention;

[0074] Figure 3 It is a flowchart of the ICP algorithm of the present invention;

[0075] Figure 4 It is an effect diagram of obstacle frame selection of the present invention;

[0076] Figure 5 It is an installation effect diagram of the rope length measurement sensor and the swing angle measurement sensor of the present invention;

[0077] Figure 6 It is a mechanical transmission structure diagram of the trolley of the present invention;

[0078] Figure 7 It is a control flowchart of the electrical control system of the present invention;

[0079] Figure 8 It is an anti-collision disposal flowchart of the present invention;

[0080] Figure 9 It is a schematic diagram of the corresponding disposal methods for anti-collision levels of the present invention;

[0081] Figure 10 It is a block diagram of the bridge crane anti-collision system in Embodiment 2 of the present invention;

[0082] Figure 11 It is a working flowchart of the bridge crane anti-collision system in Embodiment 2 of the present invention;

[0083] Figure 12 It is a physical and structure diagram of the lidar in Embodiment 3 of the present invention;

[0084] Figure 13 It is an installation schematic diagram of the lidar in Embodiment 3 of the present invention;

[0085] Figure 14 It is a design context block diagram of the safety anti-collision method for a bridge crane based on lidar SLAM in Embodiment 3 of the present invention;

[0086] Figure 15 This is the test diagram of the anti-collision disposal distance for Embodiment 3 of the present invention;

[0087] Figure 16 This is the line graph of the experimental data of the gantry crane anti-collision for Embodiment 3 of the present invention.

[0088] Among them: In Figure 5 Figure (a) is the installation effect diagram of the rope length measurement sensor, and Figure (b) is the installation schematic diagram of the rope length measurement sensor; Figure (c) is the installation schematic diagram of the swing angle measurement sensor;

[0089] In Figure 12 Figure (a) is the physical diagram of the lidar; Figure (b) is the structural diagram of the lidar;

[0090] In Figure 13 Figure (a) is the front view after the lidar is installed; Figure (b) is the bottom view after the lidar is installed; Specific implementation manner

[0091] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0092] Embodiment 1:

[0093] Please refer to Figures 1 - 16 As shown in the figure, this embodiment provides a technical solution: A safety anti-collision method for a gantry crane based on lidar, including:

[0094] Step 1: Use the lidar to obtain the point cloud data when the gantry crane is working on the bridge, and use the SLAM (Simultaneous Localization and Mapping) algorithm to process the point cloud data

[0095] This step generally uses the SLAM algorithm as the framework to implement the environmental detection function. It includes: Point cloud preprocessing: Filter and downsample the lidar point cloud to reduce the data volume and noise interference; Matching and positioning: Based on the ICP (Iterative Closest Point) algorithm, optimize the initial pose to obtain a more accurate pose estimate; Loop detection: Detect whether there is a loop by matching the point cloud between the current frame and the historical frame, and update the pose estimate; Mapping: Construct a three-dimensional map with the optimized pose and point cloud data. The specific process includes:

[0096] Step 1.1: Preprocess the point cloud data, filter and downsample the lidar point cloud

[0097] Step 1.1.1: Input the point cloud information P = {p i | i = 1, 2,..., n}, find the minimum vertex (x min , y min , z min ) and the maximum vertex (x max , y max , z max ). Use these two points as the vertices of the bounding box, establish a coordinate system with the minimum point as the origin, and enclose all the input point clouds;

[0098] Step 1.1.2: Divide the voxel cubes, set the side length of the unit cube l = (r, r, r), and calculate the number of voxel cube divisions through the following formula:

[0099]

[0100] where length is the length of the unit cube, width is the width of the unit cube, height is the height of the unit cube, and x max , y max , z max、 x min , y min , z min represent the maximum or minimum coordinate values on the x, y, and z axes respectively.

[0101] Step 1.1.3: Calculate the geometric center of each voxel cube, and replace all the points in the cube with the coordinates of the center point M(x a , y a , z a ). The center calculation formula is:

[0102]

[0103] where x centre , y centre , z centre represent the geometric center coordinate values on the x, y, and z axes respectively; x i , y i , z i represent the coordinate values of each point on the x, y, and z axes.

[0104] Step 1.1.4: Replace all the point clouds in the voxel cube with the geometric center point M(x a , y a , z a ) of the voxel cube, and integrate the geometric centers in all the cubes into new point cloud information.

[0105] In actual use, it is found that using the center to represent the points within a grid is not representative in some cases. Therefore, grid centroid optimization is carried out, and the centroid of all the point clouds within the voxel is used to represent the point cloud within the voxel grid, which has a wider application range and better sampling effect. Among them, the geometric centroid point M center The calculation formula is as follows:

[0106]

[0107] Radius filtering is to calculate the number of other point clouds within the set radius for each point cloud by setting the filtering radius. If the number of other point clouds of a point cloud is less than the set threshold, that point cloud will be removed. This filtering method can relatively simply and quickly remove the useless point clouds in space, has a good effect on different types of noise, and has a fast operation speed.

[0108] Step 1.2: Perform coordinate system conversion

[0109] After the lidar collects the reflected point cloud information, it is transmitted to the processor for processing. However, the point cloud coordinate information at this time is in the space coordinate system established with the lidar as the coordinate origin. In the environmental detection module of the anti-collision system, the point cloud information needs to be converted into the space coordinate system with the center of the suspended object as the origin in order to provide a basis for the anti-collision system to judge whether there is a collision risk between the suspended object and the obstacle.

[0110] Step 1.2.1: In the anti-collision system designed in this embodiment, the lidar is placed directly above the suspended object under the bridge crane bridge. Space coordinate systems are established with the center point O of the lidar and the center point O' of the suspended object as the coordinate origins respectively, and the directions of their respective axes are as Figure 2 shown. The system coordinate conversion method is as shown in the formula:

[0111]

[0112] Among them, the physical meaning of the matrix transformation does not need to be explained. R is the rotation matrix, and T is the translation matrix.

[0113] Step 1.2.2: Let a, β, and γ be the angles by which the coordinate system of the suspended object rotates relative to the x, y, and z axes of the lidar coordinate system, and t x , t y , t z be the translation distances of the two coordinate systems on the corresponding axes during the coordinate system conversion process. Due to the motion characteristics of the bridge crane, the coordinate system conversion in this embodiment only involves translational transformation and does not involve rotational transformation.

[0114]

[0115] Among them, the translation transformation distance in the z-axis direction can be obtained by a rope length sensor. For an object with a small volume, the translation transformation distances in the x-axis or y-axis can be regarded as 0. For an object with a large volume or a long and straight object, the translation transformation distance needs to be manually measured and calibrated according to the movement direction of the bridge crane.

[0116] Step 1.3: After completing the coordinate transformation, use the ICP algorithm for point cloud matching and positioning

[0117] Step 1.3.1: Read two frames of point cloud data obtained by the lidar, select two groups of point cloud data X and Y, and randomly initialize the transformation matrices R and t;

[0118] Step 1.3.2: Find corresponding points. For each point in X, use a distance metric (usually the Euclidean distance) to find the nearest point in B to A as the corresponding point of A.

[0119] Directly finding the nearest point has a high computational complexity. Generally, by setting a distance threshold, when the distance between points is less than a certain threshold, it is considered that the corresponding point is found. Take points in X Take points in Y If the formula is satisfied, it can be considered as the corresponding point, as shown in the following formula:

[0120]

[0121] Step 1.3.3: Optimize the rotation matrix R and the translation matrix t

[0122] When the corresponding points are obtained, use the corresponding points to estimate the rotation matrix R and the translation matrix t. The least squares method can be used to estimate and calculate the optimal matrix, and the minimization objective function E(R,t) is determined as:

[0123]

[0124] Step 1.3.4: Iteratively calculate the transformation matrices R and t

[0125] When the new transformation matrices R and t are optimized and the positions of some point clouds change, and the positions of their nearest points also change accordingly; go back to the method of finding corresponding points in Step 1.2 from here. After obtaining the new corresponding points, go to Step 1.3 again to update the rotation matrix R and the translation matrix t; Steps 1.2 and 1.3 are iterated until the designed iteration termination condition is met and the objective function is less than the convergence error. The ICP algorithm flow chart is as Figure 3 shown.

[0126] Step 1.4: After using the ICP (Iterative Closest Point) algorithm for point cloud matching and positioning in Step 1.3, by matching the point clouds between the current frame and the historical frame, detect whether there is a loop closure. After updating the pose estimation, use the optimized pose and point cloud data to construct a three-dimensional map.

[0127] Step 2: After completing the mapping in the transfer environment in Step 1, perform environmental object segmentation

[0128] After the environmental detection module completes the mapping in the transfer environment, the gantry crane needs to detect the passable areas and the objects that affect its movement in the environment. Therefore, in order to make safe and correct active anti-collision disposal, it is necessary to cluster and segment the objects in the transfer environment.

[0129] Step 2.1: Use the segmentation algorithm based on RANSAC to perform clustering segmentation on the point cloud

[0130] The core of using the segmentation algorithm based on RANSAC to perform clustering segmentation on the point cloud is to select a set of point cloud data from the overall point cloud information. This set of point cloud data needs to fit the plane model and calculate its corresponding parameters. The specific implementation method is as follows:

[0131] Step 2.1.1: First, randomly select three point clouds from all the point cloud data, judge the geometric relationship of the three point clouds. If the three points are collinear and not in the same plane, then reselect three point clouds until they are non-collinear and in the same plane, and then calculate the relevant model parameters of the plane they form.

[0132] Step 2.1.2: Traverse and calculate the distances from other point clouds to the plane model constructed in Step 2.1.1, and compare them with the set distance threshold. Store the point cloud information whose distance to the plane model is less than the distance threshold.

[0133] Step 2.1.3: Repeat Step 2.1.1 and Step 2.1.2 until the set maximum number of iterations. Select the model with the largest number of set points as the fitting plane result according to the information stored in Step 2.1.2 to complete the plane fitting.

[0134] Step 2.2: Use the method of enclosing with a rectangular box to enclose the point cloud clusters of the same color with a rectangular box, and then extract information such as the obstacle distance and centroid from them

[0135] After the point cloud is clustered and segmented by the RANSAC algorithm, the lidar point clouds from the same object can basically be clustered into a point cloud cluster. To ensure the realization of the anti-collision decision-making function, it is necessary to extract obstacle information from the clustered objects. In terms of obtaining obstacle information, in this embodiment, the method of surrounding with a rectangular box is adopted, and the point cloud cluster of the same color is surrounded by a rectangular box, and then information such as the distance and centroid of the obstacle is extracted from it. The specific implementation method is as follows:

[0136] Step 2.2.1: First, find the points in the point cloud cluster with the maximum and minimum coordinate values in the x, y, and z axes of the point cloud space scanned by the lidar. By subtracting the maximum value from the minimum value, calculate the length, width, and height dimensions of the rectangular bounding box of the obstacle, and construct the smallest rectangular bounding box that can envelope all the point clouds from the same object.

[0137] length = |x max - x min |

[0138] width = |y max - y min |

[0139] height = |z max - z min |

[0140] The above operations can obtain obstacle information through jsk_recognition_msgs::BoundingBoxArray in the ROS system. After the smallest rectangular bounding box is constructed, the obstacle distance information and the centroid position of the obstacle rectangular box can be obtained according to the information of the rectangular box.

[0141] Step 2.2.2: Obtain the centroid position of the obstacle

[0142] To obtain the centroid coordinate position of the obstacle, the centroid coordinate point (x, y, z) of the target obstacle can be obtained by calculating the average value of all point cloud coordinates on the x, y, and z axes, as shown in the following formula:

[0143]

[0144] Step 2.2.3: Obtain the obstacle distance information

[0145] The obstacle distance D can be obtained by calculating the distance from the centroid position (x, y, z) of the rectangular box to the lidar and the size of the rectangular box:

[0146]

[0147] Finally, verify the KITTI dataset in the ROS environment under the Ubuntu18.04 system, and get asFigure 4 Similar effect diagrams. The rectangular frame can completely enclose the result after point cloud clustering segmentation, from which obstacle information can be extracted to provide a basis for the anti-collision decision-making module.

[0148] Step 3: Establish an anti-collision control strategy

[0149] Step 3.1: Design the control hardware of the anti-collision system

[0150] The anti-collision decision-making module needs to be connected to sensors to obtain the status information of the gantry crane in order to make reasonable decisions. Therefore, in addition to the industrial computer, the anti-collision decision-making module should also include various sensors, such as Figure 5 shown.

[0151] (1) Rope length measurement sensor, used to measure the change of the wire rope length in real time, installed at the tail of the hoisting motor;

[0152] (2) Swing angle measurement sensor, and the swing angle measurement method is to measure the displacement of the pull rope by a wire displacement sensor and convert it into a swing angle;

[0153] (3) Mechanical transmission system

[0154] The mechanical transmission system is installed on the bridge and consists of the bridge, trolley and hoist running mechanisms. The trolley drive consists of a motor, gearbox, drive wheels and rails, etc., as Figure 6 shown. The gearbox is an important part that transmits the speed and torque of the motor to the drive wheels. The inside of the gearbox consists of a series of gears. Through the combination of gears of different sizes, the adjustment of the output speed and torque of the motor is realized to meet the operation requirements of the gantry crane trolley under different working conditions. The drive wheel is a key component of the gantry crane trolley running mechanism, responsible for contacting the track, transmitting power and making the trolley run on the track. The rail is the guiding component of the gantry crane trolley running mechanism, used to guide the trolley to run on the track and maintain its stability along the track direction.

[0155] The trolley running mechanism usually consists of a motor, gearbox, drive wheels, trolley bridge, wheel-rail system and guiding device, etc. The power part is basically the same as that of the trolley. Its wheel-rail system is the key component to support the trolley to run on the main girder, including wheel-rail, wheel-rail frame and wheel-rail connection device, etc. The wheel-rail consists of a rail and a wheel-rail frame, and the wheel-rail connection device fixes the wheel-rail on the main girder to ensure the stability and smoothness of the trolley during operation.

[0156] The hoisting mechanism of the bridge crane is one of the core components of the bridge crane, mainly used to realize the hoisting and lowering operations of the load. The electric motor converts electrical energy into mechanical energy through the transmission device, providing sufficient power to drive the hoisting mechanism for hoisting and lowering operations. The reducer is a key component that converts the high-speed rotation of the electric motor into the low-speed and high-torque output required by the hoisting mechanism. The drum is used to wind the wire rope and realize the hoisting and lowering of the goods. The wire rope is an important component connecting the drum and the hook, undertaking the task of hoisting and lowering the goods.

[0157] (4) Electrical control system

[0158] The electrical control system takes the programmable logic controller PLC as the core, and the trolley, crab and hoisting mechanism of the bridge crane are controlled by frequency converters for variable frequency control. During operation, the industrial control computer communicates with the PLC, analyzes the working state of the crane and the movement state of the load according to the received sensor information, and sends variable frequency control instructions to the PLC. After receiving the control instructions, the PLC completes the control of the corresponding motors by reading and writing the internal registers of the three frequency converters.

[0159] The system adopts the drive mode of "PLC - frequency converter - motor" to control the movement direction and speed of the bridge crane. The PLC controls the movement direction of the motor by energizing the corresponding coil relays of the motor. The speed control of the motor is obtained by the PLC calculating the control frequency and transmitting it to the frequency converter in the form of D / A. The frequency converter outputs the corresponding drive signal frequency according to the input current magnitude, and the motor performs stepless speed change movement according to the drive signal frequency. The principle is as Figure 7 shown.

[0160] Step 3.2: Design the anti-collision disposal method

[0161] The anti-collision disposal flow chart is as Figure 8 shown. Design the anti-collision control strategy based on the collision distance. This anti-collision method is to select the disposal method by calculating the collision time required for the load and the obstacle in the current movement state in real time and comparing it with the set safe anti-collision threshold.

[0162] Let the distance between the load and the obstacle in front of it at time t be D, and the anti-collision control judgment threshold formula based on the collision distance is set as:

[0163]

[0164] where, v r is the relative speed between the lifted object and the obstacle, α0 is the braking acceleration, t1 is the system response time, t2 is the braking force increasing stage, and d0 is the reserved buffer distance.

[0165] The following are the determination methods of each parameter,

[0166] (1) Determination of relative speed v r

[0167] The accuracy of obtaining the relative speed has a great impact on the functional effect of the anti-collision system of the gantry crane when facing moving obstacles. The timestamp function of the ROS system is used to calculate the relative speed.

[0168] (2) Braking acceleration α0

[0169] The general operating speed of the gantry crane is between 5 m / min and 20 m / min. According to the standards of the International Society of Mechanical Engineers, the start-stop time of the hoisting crane should not exceed 200 ms. The national standard in China stipulates that the stop time of the hoisting crane should be between 3 and 5 s. At the same time, the national standard also requires that when the load is released, the hoisting crane should be able to stop quickly, and the stop time should not exceed 2 s. In this embodiment, the braking acceleration is calculated according to a speed of 20 m / min and a stop time of 2 s, and α0 is taken as 0.17 m / s 2 。

[0170] (3) System response time t1

[0171] After the above analysis, it takes some time for the active anti-collision system of the gantry crane to obtain environmental information, evaluate the collision risk and send instructions to the braking system until the mechanical braking system starts to respond. According to the standards of the International Society of Mechanical Engineers, the start-stop time of the hoisting crane should not exceed 200 ms. In order to ensure the reliability of the safety distance, relevant information is collected, and in this embodiment, the system response time t2 is taken as 0.2 s.

[0172] (4) Braking force increase time t2

[0173] Since the 5t gantry crane relied on in the experiment of this embodiment uses an electromagnetic open-brake disc brake, after receiving the braking instruction and starting to respond, the electromagnetic coil is powered off, and the armature presses the rotor between the armature and the cover under the action of the spring force, and the motor brakes under the action of friction. Since the electromagnetic open-brake method has the characteristic of rapid response and the action of the spring force is instantaneous, t2 is taken as 0 in this embodiment.

[0174] (5) Reserved buffer distance d0

[0175] During the actual operation of the bridge crane, whether the obstacle is stationary or moving, it is necessary to maintain a reasonable safety distance from the obstacle after braking. The distance between the bridge crane and the obstacle after braking is crucial. If only the non-contact situation is considered without taking into account the safety reserve distance, it is very likely that the braking distance will be extended due to system errors, affecting the safety of loading and unloading. Considering safety and reliability, in order to leave some reaction time for manual handling, d0 should be taken larger. Considering the working requirements of the system in an environment where obstacles are stacked in a more complex manner. After consulting relevant materials and combining with the actual operation, d0 is taken as 0.6m in this embodiment.)

[0176] When D ttc > 100, the distance between the two will not collide temporarily, so no anti-collision treatment is carried out; when D ttc ≤ 0, the distance left for anti-collision treatment is very short, and primary anti-collision treatment is carried out. The threshold is set according to the maximum speed of the bridge crane and the minimum safe anti-collision distance between the load and the obstacle under the condition that the obstacle is stationary; when 0 < D ttc ≤ 60, the system performs secondary anti-collision treatment. The threshold is set according to the minimum safe anti-collision distance between the load and the obstacle under the above conditions plus a certain distance; when 60 < D ttc ≤ 100, the system outputs tertiary anti-collision treatment.

[0177] As Figure 9 shown, the green area indicates that the collision time meets the condition for issuing tertiary anti-collision treatment, and the buzzer emits an alarm signal to remind the operator to pay attention to the collision risk and observe the surrounding environment; the yellow area represents that the system performs secondary anti-collision treatment. Without manual control, the bridge crane automatically decelerates to 60% of the current speed; the red area represents that the system enters primary anti-collision treatment, and the system immediately issues a braking command.

[0178] Embodiment 2:

[0179] As Figure 7 、 Figure 10 and Figure 11 shown, different from the above Embodiment 1, the present invention provides another solution, a safety anti-collision system for a bridge crane based on lidar. This safety anti-collision system is a bridge crane anti-collision system, which is used to implement the safety anti-collision method for a bridge crane based on lidar SLAM as described in Embodiment 1, including an environment detection module, an anti-collision decision module, and a bridge crane control module; among them:

[0180] The environment detection module is a lidar, which is used to detect the loading environment during use and send the obtained point cloud map, real-time positioning, and obstacle information to the anti-collision decision module;

[0181] The anti-collision decision-making module is used to segment environmental objects from the received point cloud map, real-time positioning, and obstacle information, and make anti-collision decisions based on the target detection results and status information results;

[0182] The bridge crane control module is used to receive the anti-collision decision of the anti-collision decision-making module, convert the anti-collision decision into a control command, and perform vehicle speed control and braking control.

[0183] In a preferred embodiment, the lidar is installed directly above the lifted object under the bridge of the bridge crane, and is used to obtain the point cloud data during the operation of the bridge crane bridge, and process the point cloud data according to the LOAM algorithm described in step 1 of Embodiment 1 to obtain the point cloud map, real-time positioning, and obstacle information.

[0184] In a preferred embodiment, as Figure 10 and Figure 11 shown, the anti-collision decision-making module makes anti-collision decisions on the target detection results and status information results according to the environmental object segmentation method described in step 2 of Embodiment 1; it includes a sensor group and an industrial computer; where:

[0185] The sensor group is used to obtain the bridge crane status information and send the bridge crane status information to the industrial computer; it includes a rope length measurement sensor and a swing angle measurement sensor. The rope length sensor is installed at the tail of the hoisting motor and is used to measure the change in the wire rope length in real time; the swing angle measurement sensor is used to measure the pull rope displacement in real time and convert it into a swing angle.

[0186] The industrial computer is used to receive the bridge crane status information and the target detection results, analyze the working status of the crane and the movement status of the load according to the received sensor information, and send a frequency conversion control instruction to the bridge crane control module.

[0187] In a preferred embodiment, as Figure 7 shown, the bridge crane control module is an electrical control system. The electrical control system uses a programmable logic controller PLC as the core and adopts a "PLC-inverter-motor" drive mode to control the movement direction and speed of the bridge crane. Specifically: the PLC controls the energization of the corresponding coil relay of the motor to achieve the control of the movement direction of the motor. The speed control of the motor is obtained by the PLC calculating the control frequency and transmitting it to the inverter in the form of D / A. The inverter outputs the corresponding drive signal frequency according to the input current size, and the motor performs stepless speed change movement according to the drive signal frequency. The inverter is installed on the structures such as the trolley, crab, and hoisting mechanism of the bridge crane.

[0188] Embodiment 3:

[0189] As Figure 12 、 Figure 13 、Figure 14 , Figure 15 and Figure 16 As shown in Figure 14 , Figure 15 and Figure 16 , different from the above embodiments, to verify the effectiveness of the safety anti-collision method for gantry cranes based on lidar as described in Embodiment 1 and the gantry crane anti-collision system as described in Embodiment 2, on the basis of Embodiment 1, in this embodiment, by building a verification system as described in Figure 12 , Figure 13 , Figure 14 and Figure 15 , the above system and method are verified, and a line chart of the experimental data of gantry crane anti-collision is obtained as shown in Figure 16 . It can be seen from Figure 16 that: Figure 12 , Figure 13 and Figure 14 and Figure 15 the above system and method are verified, and a line chart of the experimental data of gantry crane anti-collision is obtained as shown in Figure 16 . It can be seen from Figure 16 that: Figure 16 Through Figure 16 it can be seen that:

[0190] (1) The experimental results under the condition that the gantry crane is under the rated load (2t) show that after the anti-collision function is turned on, when the gantry crane brakes at a slow speed (5m / min), the minimum distance between the lifted object and the obstacle after braking is 63cm, and when the gantry crane brakes at a fast speed (15m / min), the minimum distance between the lifted object and the obstacle after braking is 62cm. The expected reserved safety buffer distance of the safety distance model designed in Embodiment 1 of the present invention is 60cm. The experimental data show that the designed safety anti-collision model conforms to the braking process of the gantry crane, and the system can meet the safety requirements of gantry crane anti-collision.

[0191] (2) The experimental results under the condition that the gantry crane is under the rated load (500kg) show that according to the line chart, the distance between the load and the obstacle after braking in the slow state is larger than that in the fast state. No matter whether it is fast or slow, the final distance is greater than 60cm. However, in some cases, it is close to 60cm. After analysis, the reason is that the accuracy of the coordinate system transformation matrix is not enough. Therefore, it can be concluded that when the mass of the lifted object is 500kg and relatively small, the system basically meets the safety requirements of the gantry crane, but the problem of the swing amplitude of the lifted object needs to be further analyzed in depth.

[0192] (3) It can be seen from the line chart that the distance between the load and the obstacle after braking is not affected by three different obstacles, and the distance between the load and the obstacle after braking in the slow state is larger than that in the fast state. The experimental results show that the environmental detection algorithm adopted in Embodiment 1 of the present invention can accurately detect environmental information, and the obstacle target detection method can correctly obtain information on various obstacles, and the accuracy meets the anti-collision requirements of the gantry crane.

[0193] As can be seen from the above results, the method described in Embodiment 1 of the present invention uses lidar SLAM technology to achieve autonomous positioning and environmental mapping of the load transfer of bridge cranes. After obtaining the point cloud information and establishing the environmental map, the point cloud segmentation method is used to segment the ground obstacles and establish rectangular bounding boxes for collision detection. After obtaining the obstacle information, the anti-collision level is judged according to the state of the load itself (volume, shape, speed), and then corresponding anti-collision disposal actions are taken, which can effectively meet the safety anti-collision requirements of bridge cranes; it solves the problems existing in traditional bridge cranes, such as insufficient environmental detection ability, insufficient protection against load hoisting collision, and inability to complete its own position positioning and surrounding environmental map construction.

[0194] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A safety anti-collision method for bridge crane based on laser radar, characterized in that: include: Step 1: Obtain point cloud data of the bridge crane during operation through laser radar, and process the point cloud data using SLAM algorithm; Step 2: After completing the mapping in the reprinted environment in step 1, perform environmental object segmentation Step 2.1: Use the RANSAC-based segmentation algorithm to cluster and segment the point cloud; Step 2.2: Use the rectangular frame enclosing method to extract obstacle information; Step 3: Build a collision avoidance control strategy The collision avoidance control judgment threshold based on the collision distance is: Among them, v r is the relative speed between the load and the obstacle, α0 is the braking acceleration, t1 is the system response time, t2 is the braking force increase stage, d0 is the reserved buffer distance, and D is the distance between the load and the obstacle in front at time t.

2. A laser radar based bridge crane safety anti-collision method as claimed in claim 1, characterized in that: Step 1 The process of processing point cloud data using SLAM algorithm includes: Step 1.1: Preprocessing of point cloud data: filtering and downsampling of LiDAR point cloud; Step 1.2: Perform coordinate system conversion to convert the point cloud information into a spatial coordinate system with the center of the hanging object as the origin; Step 1.3: After completing the coordinate transformation, use the ICP algorithm to perform point cloud matching and positioning; Step 1.4: After using the ICP algorithm to perform point cloud matching and positioning in step 1.3, detect whether there is a loop by matching the point cloud between the current frame and the historical frame. After updating the pose estimation, use the optimized pose and point cloud data to build a three-dimensional map.

3. A laser radar based bridge crane safety anti-collision method as claimed in claim 1, characterized in that: Step 1.1 The process of point cloud data preprocessing includes: Step 1.1.1: Input the point cloud information to be processed P = {p i |i=1,2,...,n}, find the minimum vertex (x min ,y min ,z min ) and the maximum vertex (x max ,y max ,z max ), use these two points as the vertices of the bounding box, and establish a coordinate system with the minimum point as the origin to enclose all the input point clouds; Step 1.1.2: Divide the voxel cube, set the side length of the unit cube to l = (r, r, r), and calculate the number of voxel cube divisions by the following formula: Among them, length is the length of the unit cube, width is the width of the unit cube, height is the height of the unit cube, x max ,y max 、z max、 x min ,y min 、z min Respectively represent the maximum or minimum coordinate value on the x, y, and z axes; Step 1.1.3: Calculate the geometric center of each voxel cube, with the center point M(x a ,y a ,z a ) coordinates to replace all points in the cube, and the center calculation formula is: Among them, x centre ,y centre , z centre Respectively represent the geometric center coordinates on the x, y, and z axes; x i ,y i , z i Represents the coordinate values ​​of each point on the x, y, and z axes; Step 1.1.4: Take the geometric center point M(x a ,y a ,z a ) replaces all point clouds in the cube and integrates all geometric centers in the cube into new point cloud information.

4. A laser radar based bridge crane safety anti-collision method as claimed in claim 1, characterized in that: Step 1.3 The process of using the ICP algorithm to perform point cloud matching and positioning includes: Step 1.3.1: Read two frames of point cloud data acquired by the laser radar, select two sets of point cloud data X and Y, and randomly initialize the transformation matrix R and t; Step 1.3.2: Find corresponding points. For each point in the point cloud data X, use the distance metric to find the nearest point in B to A as the corresponding point of A. Step 1.3.3: Optimize the rotation matrix R and translation matrix t After obtaining the corresponding points, the corresponding points are used to estimate the rotation matrix R and the translation matrix t; specifically, the least squares method is used to estimate and calculate the optimal matrix, and the minimized objective function E(R, t) is determined as: Step 1.3.4: Iterate the transformation matrix R and t When the new transformation matrix R and t are obtained through optimization, the positions of some point clouds change, and the positions of their nearest points also change accordingly; return to the method of finding corresponding points in step 1.2, and then go to step 1.3 after obtaining new corresponding points, and update the rotation matrix R and translation matrix t again; steps 1.2 and 1.3 are iterated until the designed iteration termination conditions are met and the objective function is less than the convergence error.

5. A laser radar based bridge crane safety anti-collision method as claimed in claim 1, characterized in that: Step 2.1 The process of clustering and segmenting the point cloud using the RANSAC-based segmentation algorithm includes: Step 2.1.1: First, select three point clouds from all the point cloud data and determine the geometric relationship between the three point clouds. If the three points are collinear, they are not in the same plane. Then, reselect the three point clouds until they are not collinear and are in the same plane, and then calculate the relevant model parameters of the plane they constitute; Step 2.1.2: traverse and calculate the distance from other point clouds to the plane model constructed in step 2.1.1, compare it with the set distance threshold, and store the point cloud information whose distance to the plane model is less than the distance threshold; Step 2.1.3: Repeat steps 2.1.1 and 2.1.2 until the maximum number of iterations is set. According to the information stored in step 2.1.2, the model with the most set points is selected as the plane fitting result to complete the plane fitting.

6. A laser radar based bridge crane safety anti-collision method as claimed in claim 1, characterized in that: Step 2.2 uses the rectangular frame enclosing method to extract obstacle information, and the process includes: Step 2.2.1: First, find the point cloud with the largest and smallest coordinate values ​​in the x, y, and z axes of the point cloud space scanned by the lidar in the point cloud cluster, and subtract the maximum and minimum values ​​to find the length, width, and height of the obstacle rectangular bounding box, and construct the minimum rectangular bounding box that can enclose all point clouds from the same object: length=|x max -x min | width=|y max -y min |; height=|z max -z min | Step 2.2.2: Get the centroid position of the obstacle The centroid coordinates (x, y, z) of the target obstacle are obtained by calculating the average value of all point cloud coordinates on the x, y, and z axes: Step 2.2.3: Obtain obstacle distance information The obstacle distance information D is calculated based on the distance from the center position (x, y, z) of the rectangular frame to the laser radar and the size of the rectangular frame:

7. A safety anti-collision system for bridge crane based on laser radar, characterized in that: The safety anti-collision system is a bridge crane anti-collision system, and the bridge crane anti-collision system is used to implement the bridge crane safety anti-collision method based on laser radar SLAM as described in any one of claims 1-6.

8. The laser radar-based bridge crane safety anti-collision system according to claim 7, characterized in that: include: The environment detection module is used to detect the reloading environment when in use, and send the obtained point cloud map, real-time positioning and obstacle information to the anti-collision decision module; The anti-collision decision module is used to segment environmental objects based on the received point cloud map, real-time positioning and obstacle information, and make anti-collision decisions based on the target detection results and status information results; The bridge crane control module is used to receive the anti-collision decision of the anti-collision decision module and convert the anti-collision decision into a control command to perform vehicle speed control and braking control.

9. A laser radar based bridge crane safety anti-collision system as claimed in claim 8, characterized in that: The anti-collision decision module comprises: The sensor group is used to obtain the bridge crane status information and send the bridge crane status information to the industrial computer; it includes a rope length measurement sensor and a swing angle measurement sensor. The rope length sensor is installed at the tail of the lifting motor and is used to measure the length change of the wire rope in real time; the swing angle measurement sensor is used to measure the displacement of the pull rope in real time and convert it into a swing angle; The industrial computer is used to receive the bridge crane status information and target detection results, analyze the working status of the crane and the movement status of the load according to the received sensor information, and send frequency conversion control instructions to the bridge crane control module.

10. A laser radar based bridge crane safety anti-collision system as claimed in claim 8, characterized in that: The bridge crane control module is an electrical control system, which is based on a programmable controller PLC and controls the movement direction and speed of the bridge crane based on the PLC-inverter-motor driving method.

Images