Crane laser radar anti-collision system and method based on collision time grading protection

Abstract

The application provides a crane laser radar anti-collision system and method based on collision time grading protection, which comprises: a laser radar sensing module installed on the main structure of the crane, used for scanning the working area and obtaining three-dimensional point cloud data of obstacles; a main control processing module, used for pre-processing the point cloud data to identify obstacles, calculating the relative speed and collision time according to the real-time running speed and displacement of the crane, and dynamically dividing the protection zone and outputting the grading control instruction according to the comparison result of the collision time and the preset multi-level threshold; a communication module, used for data transmission; a redundant safety execution module, adopting a double-channel safety relay structure and being arranged in parallel with the main control processing module, when the communication interruption of the main control processing module, software deadlock or emergency stop signal triggering is detected, forcibly cutting off the power circuit and triggering the mechanical brake, realizing the fault safety self-locking. The application realizes non-contact high-precision anti-collision and hardware-level forced braking.

Description

Technical Field

[0001] This invention belongs to the field of crane technology, specifically relating to a crane laser radar anti-collision system and method based on collision time-based graded protection. Background Technology

[0002] Traditional crane collision avoidance systems typically employ mechanical limit switches or contact sensors.

[0003] Mechanical limit switches are triggered by physical collisions, which can lead to mechanical wear and tear over time, resulting in a gradual decrease in positioning accuracy and an inability to provide non-contact early warning. Human visual inspection has blind spots and is prone to fatigue-related misjudgments, while ultrasonic and radar systems are susceptible to interference from dust, rain, and fog.

[0004] Existing electronic collision avoidance systems mostly use a single distance threshold, which lacks the ability to predict the movement trend of obstacles, making them prone to false triggering or insufficient braking.

[0005] Therefore, there is an urgent need for a collision avoidance system with non-contact, high-precision sensing and motion trend prediction capabilities. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a crane lidar collision avoidance system and method with non-contact high-precision sensing and motion trend prediction based on collision time-based hierarchical protection.

[0007] The technical solution of this invention is: a crane lidar anti-collision system based on collision time-based graded protection, comprising:

[0008] The lidar sensing module is installed on the main structure of the crane and is configured to scan the working area and acquire three-dimensional point cloud data of obstacles.

[0009] The main control processing module is communicatively connected to the lidar sensing module and the crane drive mechanism. It is configured to preprocess lidar point cloud data to identify obstacles, calculate relative speed and collision time based on the crane's real-time running speed and displacement, and dynamically divide multi-level protection zones and output corresponding hierarchical control commands based on the comparison results of collision time and preset multi-level thresholds.

[0010] The communication module is used for data transmission between the lidar sensing module and the main control processing module.

[0011] The redundant safety execution module adopts a dual-channel safety relay architecture and is connected in parallel with the main control processing module. The input terminal of the safety relay receives the emergency stop signal from the main control processing module and the external hard-wired emergency stop signal, and the output terminal is directly connected in series in the power circuit contactor coil and brake coil circuit of the crane drive motor. The redundant safety execution module is configured to forcibly cut off the power supply to the power circuit and trigger the mechanical brake when it detects a communication interruption, software deadlock, or emergency stop signal trigger of the main control processing module, thereby achieving fault-safe self-locking.

[0012] Furthermore, the main control processing module is equipped with a virtual fence logic unit:

[0013] The virtual fence logic unit dynamically divides the work area into a safe zone, a warning zone, a deceleration zone, and an emergency stop zone based on the collision time and a preset multi-level threshold.

[0014] In the safe zone, if the collision time is greater than the first threshold, the system maintains normal operation and does not output intervention commands.

[0015] In the warning zone, when the collision time is greater than the second threshold and less than or equal to the first threshold, an audible and visual alarm signal is output.

[0016] In the deceleration zone, when the collision time is greater than the third threshold and less than or equal to the second threshold, a power cut-off command is output, and inertial gliding is used to trigger electric braking.

[0017] In the emergency stop zone, when the collision time is less than or equal to the third threshold, a dual-channel emergency stop command is output to control the combined braking of the electric brake and the mechanical brake.

[0018] Wherein, the first threshold is greater than the second threshold, and the second threshold is greater than the third threshold.

[0019] Furthermore, the main control processing module includes:

[0020] The point cloud denoising unit uses a statistical filtering algorithm to calculate the average k-nearest neighbor distance for each point and remove isolated noise points that exceed the standard deviation of the global mean by a preset multiple.

[0021] The point cloud clustering unit uses Euclidean distance clustering algorithm or voxel grid segmentation algorithm to segment spatially adjacent point clouds into independent obstacle clusters and remove noise clusters with fewer than a preset threshold number of points.

[0022] The obstacle centroid extraction unit extracts the centroid coordinates of each obstacle cluster as the obstacle location feature.

[0023] Furthermore, before the point cloud denoising unit uses a statistical filtering algorithm to remove isolated noise points, the main control processing module first performs voxel grid downsampling on the point cloud data.

[0024] In the voxel grid downsampling, the three-dimensional working space is divided into several voxel grids with preset side lengths. The centroid of all point clouds in a single voxel grid is taken as the representative point of that voxel grid to complete the point cloud compression.

[0025] Furthermore, in the Euclidean distance clustering algorithm, point clouds with spatial spacing less than a preset clustering distance threshold are grouped into the same obstacle cluster, and obstacle clusters with a number of point clouds less than a preset number of points are filtered out.

[0026] Furthermore, the main control processing module includes:

[0027] The relative distance calculation unit calculates the relative distance between the crane and the obstacle based on the obstacle's center coordinates and the real-time displacement fed back by the crane's encoder.

[0028] The relative velocity calculation unit collects a series of relative distance sequences and corresponding timestamps from multiple past frames, and obtains the relative velocity by fitting the relationship between relative distance and time through linear regression.

[0029] The collision time calculation unit calculates the collision time by taking the ratio of the current relative distance to the absolute value of the relative speed when the relative speed is negative, i.e., when the obstacle is approaching the crane; when the relative speed is greater than or equal to zero, it is determined that there is no risk of collision.

[0030] Furthermore, the dual-channel safety relay architecture of the redundant safety execution module is configured as follows:

[0031] During normal operation, cross-validation is performed on both channels;

[0032] When any channel experiences contact adhesion, wire breakage, or inconsistent input signals, the monitoring circuit immediately cuts off the power supply to all output coils and forcibly disconnects the main contacts using a spring reset mechanism.

[0033] When the main control processing module fails to send a heartbeat maintenance signal within the specified period, the safety relay enters a de-energized state and forcibly disconnects the output.

[0034] Furthermore, the lidar sensing module includes multiple lidars, which are installed at different structural positions on the crane; the main control processing module is configured to stitch together the point cloud data of the multiple lidars to generate a full-view point cloud data.

[0035] Furthermore, the lidar sensing module includes at least one single-line lidar or a multi-line lidar;

[0036] The scanning frequency of the lidar sensing module is greater than or equal to 10Hz, and the detection distance range is from 0.1m to 50m.

[0037] The lidar sensing module is fixedly installed on the crane's main trolley or auxiliary trolley traveling mechanism and moves synchronously with the crane.

[0038] The crane lidar collision avoidance method based on collision time-based graded protection, employing the crane lidar collision avoidance system described above, includes the following steps:

[0039] The lidar sensing module periodically scans and acquires 3D point cloud data of the work area, which is then uploaded to the main control processing module via the communication module.

[0040] The main control processing module preprocesses the point cloud data, removes environmental noise, identifies obstacle clusters, and extracts obstacle locations;

[0041] The main control processing module calculates the relative speed and collision time between the crane and the obstacle based on the real-time operating speed and displacement fed back by the crane encoder.

[0042] The main control processing module dynamically classifies the current protection level based on the comparison results between the collision time and the preset multi-level thresholds, generates corresponding hierarchical control commands, and outputs them.

[0043] The redundant safety execution module monitors the communication status and heartbeat signal of the main control processing module in real time. When it detects that the main control processing module has a communication interruption, software deadlock, or receives an emergency stop signal, it bypasses the software logic to directly cut off the power supply to the power circuit and trigger the mechanical brake, so that the system enters a fault-safe self-locking state.

[0044] The beneficial effects of this invention are:

[0045] (1) In this invention, the lidar sensing module actively scans the work area in a non-contact manner, is not affected by mechanical wear, and can achieve centimeter-level ranging accuracy, which fundamentally solves the problem of the decrease in positioning accuracy caused by long-term physical collision of traditional mechanical limit switches; at the same time, the active scanning characteristics of lidar are not restricted by light conditions, effectively overcoming the blind spots and fatigue misjudgment risks of manual visual inspection.

[0046] (2) The main control processing module provides a decision basis for graded braking through point cloud preprocessing and collision time calculation. It dynamically adjusts the safety strategy through the collision time value, rather than a single threshold. It only provides warnings and does not intervene at long distances, and decelerates in stages at close distances, and finally stops suddenly, which greatly improves the efficiency and smoothness of operation.

[0047] (3) The redundant safety execution module is set up in parallel with the main control processing module. When the software logic is normal, it responds to the emergency stop command of the main control processing module. When the main control processing module is interrupted or the software is deadlocked, it bypasses the software logic and directly triggers the hardware-level braking, thus realizing the dual protection of software control and hardware protection. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the crane lidar anti-collision system based on collision time-based graded protection in this invention.

[0049] Figure 2 This is a circuit diagram of the redundant safety execution module in this invention.

[0050] Figure 3 This is a flowchart of the crane lidar anti-collision method based on collision time-based graded protection in this invention. Detailed Implementation

[0051] Various exemplary embodiments of the invention will now be described in detail with reference to the accompanying drawings. The descriptions of the exemplary embodiments are merely illustrative and are in no way intended to limit the invention or its application or use. The invention can be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to make the invention thorough and complete, and to fully express the scope of the invention to those skilled in the art. It should be noted that, unless otherwise specifically stated, the relative arrangement of components and steps, the composition of materials, numerical expressions, and values ​​set forth in these embodiments should be interpreted as merely exemplary and not as limiting.

[0052] The terms "first," "second," and similar words used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different parts. Words such as "including" or "comprising" mean that the element preceding the word encompasses the element listed after it, without excluding the possibility of encompassing other elements. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0053] like Figure 1 As shown, a crane lidar collision avoidance system based on collision time-graded protection is disclosed, comprising:

[0054] The lidar sensing module 1 is installed on the main structure of the crane and is configured to scan the working area and acquire three-dimensional point cloud data of obstacles.

[0055] The main control processing module 2 is connected to the lidar sensing module 1 and the crane drive mechanism 5. It is configured to preprocess the lidar point cloud data to identify obstacles, calculate the relative speed and collision time based on the crane's real-time running speed and displacement, and dynamically divide the multi-level protection zone and output the corresponding hierarchical control commands based on the comparison results of the collision time and the preset multi-level thresholds.

[0056] Communication module 3 is used for data transmission between lidar sensing module 1 and main control processing module 2;

[0057] The redundant safety execution module 4 adopts a dual-channel safety relay architecture and is set in parallel with the main control processing module 2. The input terminal of the safety relay receives the emergency stop signal from the main control processing module 2 and the external hard-wired emergency stop signal, and the output terminal is directly connected in series in the power circuit contactor coil and brake coil circuit of the crane drive motor. The redundant safety execution module 4 is configured to forcibly cut off the power supply to the power circuit and trigger the mechanical brake when it detects that the main control processing module 2 has a communication interruption, software deadlock or emergency stop signal is triggered, so as to achieve fault safety self-locking.

[0058] In this embodiment, the lidar sensing module 1 actively scans the work area in a non-contact manner, unaffected by mechanical wear, achieving centimeter-level ranging accuracy. This fundamentally solves the problem of decreased positioning accuracy caused by long-term physical collisions in traditional mechanical limit switches. Simultaneously, the lidar's active scanning characteristics are unaffected by lighting conditions, effectively overcoming blind spots and fatigue-related misjudgments inherent in manual visual inspection. The main control processing module 2 provides a decision-making basis for graded braking through point cloud preprocessing and collision time calculation. It dynamically adjusts the safety strategy based on collision time values, rather than a single threshold. At long distances, it only provides warnings without intervention; at close distances, it performs graded deceleration before finally stopping, greatly improving work efficiency and smoothness. The redundant safety execution module 4 is connected in parallel with the main control processing module 2. When the software logic is normal, it responds to the emergency stop command from the main control processing module 2. When communication with the main control processing module 2 is interrupted or the software is deadlocked, it bypasses the software logic and directly triggers hardware-level braking, achieving dual protection through software control and hardware protection.

[0059] Specifically, the lidar sensing module 1 uses a single-line lidar with a scanning frequency of 10Hz, which is fixedly installed at the front end of the crane's trolley traveling mechanism and moves synchronously with the crane. The main control processing module 2 uses a PLC controller and communicates with the lidar via Ethernet. The redundant safety execution module 4 uses a dual-channel safety relay, whose output contacts are connected in series in the contactor coil of the crane's traveling motor power circuit and the hydraulic brake coil circuit. For example, the redundant safety execution module 4 uses a SICK RLY3-MULT1 type dual-channel safety relay. During normal crane operation, the lidar outputs point cloud frames in real time at a frequency of 10Hz, which are processed by the main control module to generate hierarchical control commands. When the PLC experiences a watchdog timeout crash, the safety relay automatically disconnects its output within 20ms after detecting the loss of the heartbeat signal, cutting off the power circuit and triggering the brake.

[0060] In some embodiments, the main control processing module 2 is configured with a virtual fence logic unit:

[0061] The virtual fence logic unit dynamically divides the work area into a safe zone, a warning zone, a deceleration zone, and an emergency stop zone based on the collision time and a preset multi-level threshold.

[0062] In the safe zone, where the collision time exceeds the first threshold, the system maintains normal operation and does not output intervention commands.

[0063] In the warning zone, when the collision time is greater than the second threshold and less than or equal to the first threshold, an audible and visual alarm signal is output.

[0064] In the deceleration zone, where the collision time is greater than the third threshold and less than or equal to the second threshold, a power cut-off command is output, and the vehicle uses inertial gliding to trigger electric braking.

[0065] In the emergency stop zone, where the collision time is less than or equal to the third threshold, a dual-channel emergency stop command is output to control the combined braking of the electric brake and the mechanical brake.

[0066] Among them, the first threshold is greater than the second threshold, and the second threshold is greater than the third threshold.

[0067] In this embodiment, the collision time simultaneously reflects the distance and relative speed of the obstacle, enabling a proactive assessment of collision risk. When the collision time is within the safe zone (where the collision time is relatively long), the system does not intervene in the operation, ensuring operational continuity. When the collision time enters the warning zone, only an audible and visual alarm is output to alert the operator without stopping the machine, avoiding unnecessary work interruptions. When the collision time further decreases to the deceleration zone, power is cut off, inertial gliding is utilized, and electric braking is triggered to achieve smooth deceleration rather than sudden braking, reducing mechanical impact and equipment wear. Only when the collision time reaches the emergency stop threshold is the combined electric and mechanical brakes triggered to ensure a safe stop with maximum braking torque.

[0068] As an example, the first threshold is set to 8 seconds, the second threshold to 5 seconds, and the third threshold to 3 seconds. Assuming the crane is approaching an obstacle at a speed of 2 m / s and the current collision time is 6 seconds, the system determines that it has entered the warning zone and outputs an audible and visual alarm signal. The driver, aware of the risk, can proactively decelerate or adjust the path. If the collision time drops to 4 seconds, it enters the deceleration zone. The main control module automatically cuts off the motor power circuit, the contactor disconnects, and the crane decelerates by inertia, while the electric brake is engaged. If the collision time continues to drop to 2.5 seconds, it enters the emergency stop zone. The main control module outputs a dual-channel emergency stop command to the safety relay. The safety relay contacts disconnect, the power supply to the power circuit is cut off, and the electric and mechanical brakes activate simultaneously, stopping the crane within the shortest distance.

[0069] In some embodiments, the main control processing module 2 includes:

[0070] The point cloud denoising unit uses a statistical filtering algorithm to calculate the average k-nearest neighbor distance for each point and remove isolated noise points that exceed a preset multiple of the global mean standard deviation; preferably, it removes isolated noise points that exceed twice the global mean standard deviation.

[0071] The point cloud clustering unit uses either the Euclidean distance clustering algorithm or the voxel grid segmentation algorithm to segment spatially adjacent point clouds into independent obstacle clusters and remove noise clusters with fewer than a preset threshold. The voxel grid segmentation algorithm divides the point cloud space into regular voxel grids and performs clustering based on the point cloud distribution within the voxels.

[0072] The obstacle centroid extraction unit extracts the centroid coordinates of each obstacle cluster as the obstacle location feature.

[0073] This embodiment achieves rapid and accurate extraction of obstacle location features from massive amounts of raw point clouds through the cooperation of a point cloud denoising unit, a point cloud clustering unit, and an obstacle centroid extraction unit. The point cloud denoising unit employs a statistical filtering algorithm, calculating the average k-nearest neighbor distance for each point and comparing it with global statistical features to effectively eliminate isolated noise points caused by dust, rain, fog, or sensor noise, thus preventing false obstacles from triggering the collision avoidance system. The point cloud clustering unit uses an Euclidean distance clustering algorithm or a voxel grid segmentation algorithm to group spatially adjacent point clouds into the same obstacle cluster, achieving accurate differentiation between different obstacles. The obstacle centroid extraction unit calculates the centroid coordinates of each cluster, compressing hundreds or thousands of point cloud data points into a single representative point coordinate, significantly reducing the data dimensionality of subsequent motion calculations.

[0074] Taking a frame of point cloud acquired by a 10Hz LiDAR as an example, the point cloud denoising unit first calculates the average distance of each point to its 10 neighboring points. If the average distance exceeds twice the standard deviation of the global mean, the point is identified as noise and removed, effectively filtering out suspended noise caused by dust. Then, the point cloud clustering unit uses the Euclidean distance clustering algorithm, setting the clustering distance threshold to 0.5m. Points with a distance less than 0.5m are grouped into the same cluster, and small clusters with fewer than 10 points are removed, resulting in several independent obstacle clusters. Finally, the obstacle centroid extraction unit calculates the arithmetic mean of the coordinates of all point cloud points in each cluster to obtain the centroid coordinates of each obstacle, which are then output as the obstacle's positional feature to the subsequent collision time calculation unit.

[0075] In some embodiments, before the point cloud denoising unit uses a statistical filtering algorithm to remove isolated noise points, the main control processing module first performs voxel grid downsampling on the point cloud data.

[0076] In voxel mesh downsampling, the 3D working space is divided into several voxel meshes with preset side lengths. The centroid of all point clouds in a single voxel mesh is taken as the representative point of that voxel mesh to complete point cloud compression.

[0077] This embodiment adds a voxel grid downsampling step before statistical filtering as a pre-processing compression method for point cloud preprocessing. By dividing the 3D working space into several voxel grids with preset side lengths, and using the centroid of all point clouds within each voxel grid as the representative point, efficient compression of point cloud data is achieved. This downsampling step significantly reduces the amount of input data for subsequent algorithms such as statistical filtering and Euclidean clustering, reducing computational resource consumption and processing time. Furthermore, the spatially uniform sampling characteristics of the voxel grid preserve the overall distribution contour of obstacles in space, without affecting the accuracy of subsequent obstacle recognition and collision time calculation.

[0078] As an example, the LiDAR outputs approximately 100,000 3D point cloud data points per frame. The main control processing module first divides the entire scanning area into a voxel grid with a side length of 0.1m. For each voxel grid into which a point cloud falls, the average coordinates of all point cloud points within that grid are taken as the representative point of that voxel grid, outputting a simplified point cloud set (approximately 5000 points). The data volume is compressed to 5% of the original, and the computation time for subsequent statistical filtering and Euclidean clustering is reduced from approximately 50ms to approximately 2ms. The 0.1m voxel size is obtained through comprehensive optimization based on the LiDAR angular resolution, crane operating speed, and the positioning accuracy required for collision avoidance: if the voxel is too large (e.g., >0.5m), small obstacles such as personnel and small equipment may be downsampled and masked, leading to missed detection; if the voxel is too small (e.g., <0.05m), the data compression effect is limited, making it difficult to meet the real-time requirements of a 10Hz LiDAR. 0.1m strikes a balance between data compression rate and obstacle detection accuracy.

[0079] Specifically, in the Euclidean distance clustering algorithm, point clouds with spatial spacing less than a preset clustering distance threshold are grouped into the same obstacle cluster, and obstacle clusters with a number of point clouds less than a preset number of points are filtered out.

[0080] As an example, the downsampled point cloud is input to the point cloud clustering unit, and the Euclidean distance clustering algorithm is used. The main control processing module 2 traverses all point clouds, setting the clustering distance threshold to 0.5m. That is, when the three-dimensional spatial Euclidean distance between any two point clouds is less than 0.5m, they are classified into the same obstacle cluster. Suppose the current scanned scene contains two adjacent obstacles: a container generating approximately 300 point clouds and a person standing 1.2m away generating approximately 15 point clouds. Since the distance between the point clouds of the container is less than 0.5m, the container is classified into one cluster; the distance between the point clouds of the person is also less than 0.5m, so the person is classified into another cluster. The minimum distance between the two clusters is 1.2m, so they are correctly distinguished as two independent obstacle clusters. Subsequently, clusters with fewer than 10 points are filtered out, and the effective obstacle clusters and their contained point cloud sets are output.

[0081] In some embodiments, the main control processing module 2 includes:

[0082] The relative distance calculation unit calculates the relative distance between the crane and the obstacle based on the obstacle's center coordinates and the real-time displacement fed back by the crane's encoder.

[0083] The relative velocity calculation unit collects a series of relative distance sequences and corresponding timestamps from multiple past frames, and obtains the relative velocity by fitting the relationship between relative distance and time through linear regression.

[0084] The collision time calculation unit calculates the collision time by taking the ratio of the current relative distance to the absolute value of the relative speed when the relative speed is negative, i.e., when the obstacle is approaching the crane; when the relative speed is greater than or equal to zero, it is determined that there is no risk of collision.

[0085] This embodiment achieves real-time and accurate collision time calculation through the cooperation of a relative distance calculation unit, a relative velocity calculation unit, and a collision time calculation unit. The relative distance calculation unit fuses the obstacle's center coordinates identified by the LiDAR with the real-time displacement fed back by the crane's encoder to obtain the true physical distance between the crane and the obstacle, combining sensor data with the crane's own motion state. The relative velocity calculation unit collects relative distance sequences from multiple past frames and fits the distance-time curve using linear regression. Compared to the simple two-frame difference method, linear regression utilizes statistical information from multiple frames of historical data, effectively suppressing the interference of single-frame measurement noise on velocity estimation, making velocity calculation more stable and reliable. The collision time calculation unit performs branch judgment based on the relative velocity direction: a limited collision time value is calculated only when the relative velocity is negative; when the relative velocity is zero or positive, no collision risk is determined, avoiding unnecessary intervention in risk-free scenarios.

[0086] In this embodiment, the main control processing module calculates the collision time between the crane and the obstacle according to the following steps.

[0087] 1. Relative distance calculation

[0088] Let the current position coordinates of the crane be (X) c , Y c The displacement Sc is obtained by converting the real-time displacement Sc fed back by the crane encoder; the centroid coordinates of the obstacle cluster are (X... o , Y o The horizontal relative distance D between the crane and the obstacle, obtained from the aforementioned point cloud clustering and centroid extraction, is:

[0089]

[0090] If the risk of collision in the vertical direction needs to be considered, it can be expanded to three-dimensional spatial distance:

[0091]

[0092] In this embodiment, since both the crane's trolley and gantry traveling mechanisms move in the horizontal plane and the laser radar scanning plane is parallel to the ground, the horizontal relative distance is mainly used for collision time calculation.

[0093] 2. Relative velocity calculation

[0094] Assuming the obstacle maintains a constant speed for a short period, the relative distance sequence {D1, D2, …, D} of the past N frames is collected. N} and the corresponding timestamps {t1, t2, …, t N}, t i For the acquisition time of the i-th frame, the relationship between the relative distance and time is fitted using linear regression:

[0095]

[0096] in, The value represents the relative velocity (unit: m / s), and a negative value indicates that the obstacle is approaching the crane; This represents the initial relative distance.

[0097] relative speed The calculation formula is:

[0098]

[0099] In this embodiment, the lidar scanning frequency is 10Hz, and N=5, corresponding to a time window of 0.5 seconds. This window length balances the stability of speed estimation and the sensitivity to crane acceleration and deceleration. If the window is too short, it is easily affected by single-frame measurement noise, causing speed estimation to jump; if the window is too long, it cannot respond to the crane's deceleration or steering operations in a timely manner.

[0100] 3. Collision Time-to-Collapse (TTC) Calculation

[0101] When relative velocity When an obstacle approaches the crane, the collision time TTC is the ratio of the current relative distance D to the absolute value of the relative velocity.

[0102]

[0103] When relative velocity When the obstacle is stationary or far away from the crane, and it is determined that there is no risk of collision, the TTC is set to infinity, and the system does not output an intervention command.

[0104] In some embodiments, the dual-channel safety relay architecture of the redundant safety execution module is configured as follows:

[0105] During normal operation, cross-validation is performed on both channels;

[0106] When any channel experiences contact adhesion, wire breakage, or inconsistent input signals, the monitoring circuit immediately cuts off the power supply to all output coils and forcibly disconnects the main contacts using a spring reset mechanism.

[0107] When the main control processing module fails to send a heartbeat maintenance signal within the specified period, the safety relay enters a de-energized state and forcibly disconnects the output.

[0108] Specifically, such as Figure 2 As shown, the redundant safety execution module uses SICK RLY3-MULT1 safety relays conforming to IEC 61508 to construct a dual-channel redundant architecture.

[0109] During normal operation, the redundant input circuits inside the safety relays (corresponding to the normally closed contacts k1, k2, k3 connected in parallel in the schematic diagram and their series-connected protective elements F1-F3) continuously receive enable signals from the main control processing module or external emergency stop buttons. When all input circuits are closed and in the same state, the internal logic confirms that the safety conditions are met, thereby activating the output relays K1, K2, K3 to connect the power supply to the crane's power circuit and the brake coil.

[0110] This safety relay incorporates a cross-monitoring and diagnostic circuit, corresponding to the internal logic and State indicator light driver circuit in the schematic diagram, to monitor the input, output, and internal channel status in real time. The relay will immediately enter a safe state when any of the following faults occur:

[0111] Contact adhesion or input circuit failure: If any input channel experiences contact adhesion or line breakage, the monitoring circuit will detect inconsistencies in the states of multiple input signals, immediately cut off the power supply to the coils of all output relays, and forcibly disconnect the output contacts K1, K2, and K3 by resetting the relays.

[0112] Output circuit fault: If any output contact becomes stuck, the relay can detect the fault through internal diagnostics and lock in the safe disconnect state during the next power-on self-test or periodic test.

[0113] Main control unit failure: When the main control processing module crashes, the watchdog resets, or communication with the lidar times out, its safety output contact or dedicated safety output module will disconnect, causing the input circuit of the safety relay to lose power. Based on the fail-safe principle, the safety relay output contact immediately disconnects, bypassing any potentially faulty software logic and directly physically cutting off the crane's main power supply and brake holding circuit, achieving a hardware-level emergency stop.

[0114] This embodiment integrates signal evaluation, logic processing, and safety output through a dedicated safety relay. Its internal dual-channel redundancy, cross-checking, and diagnostic functions ensure that the system can reliably enter a safe state under any single fault condition, such as open circuit, short circuit, contact adhesion, component failure, or controller failure, fully complying with the requirements of SIL2 / SIL3 safety levels in the IEC 61508 functional safety standard.

[0115] In some embodiments, the lidar sensing module includes multiple lidars, which are installed at different structural positions on the crane; the main control processing module is configured to stitch together the point cloud data of the multiple lidars to generate point cloud data of the entire crane.

[0116] This embodiment eliminates blind spots in the field of view of a single radar by installing multiple LiDARs at different structural locations on the crane and using the main control processing module to stitch the point cloud data, generating a panoramic perception view of the working area. A single LiDAR has a limited scanning angle, and when installed at a single location on the crane, it cannot cover the area behind the structure or the near-end blind spots. By installing multiple radars at the lower ends or corners of the four corner beams of the crane, the scanning areas of each radar complement each other. The main control processing module uses the ICP algorithm to register the point cloud data of each radar to a unified coordinate system, achieving full coverage perception of the environment surrounding the crane without blind spots. The panoramic perception view provides more complete obstacle information for collision time calculation, avoiding collisions caused by undetected obstacles in blind spots.

[0117] As an example, one single-line lidar is installed at the lower end of each of the four corner beams of the crane, with a horizontal field of view of 270° for each. The four lidars cover the front left, front right, rear left, and rear right quadrants of the crane, respectively, with approximately 30° overlap between the scanning edges of adjacent lidars. The four lidars upload point cloud data to the main control processing module in real time via a CAN bus at a communication rate of 1Mbps.

[0118] In some embodiments, the lidar sensing module includes at least one single-line lidar or a multi-line lidar.

[0119] The scanning frequency of the lidar sensing module is greater than or equal to 10Hz, and the detection distance range is from 0.1m to 50m;

[0120] The lidar sensing module is fixedly installed on the crane's main trolley traveling mechanism or auxiliary trolley traveling mechanism and moves synchronously with the crane.

[0121] Specifically, the lidar used in this embodiment is the SICK LMS111, a single-line pulse type with a scanning frequency of 10Hz, an angular resolution of 0.5°, a detection distance range of 0.1m to 50m, and an IP67 protection rating. This lidar is fixedly installed under the front crossbeam of the crane's trolley traveling mechanism using four M8 bolts, moving synchronously with the trolley. The lidar scanning plane is 2.5m above the ground, installed at a downward tilt of 5°, with a horizontal scanning angle range of -45° to +225°, covering a 270° area in the direction of the trolley's movement and on both sides. Each scan outputs approximately 541 points, which translates to approximately 5410 points per second at 10Hz. The data volume is small and can be easily processed by a standard PLC.

[0122] In some embodiments, such as Figure 3 As shown, a crane lidar collision avoidance method based on collision time-based graded protection is disclosed, employing the crane lidar collision avoidance system as described in any of the above embodiments, including the following steps:

[0123] S1, the lidar sensing module periodically scans and acquires 3D point cloud data of the work area, and uploads it to the main control processing module through the communication module;

[0124] S2, the main control processing module preprocesses the point cloud data, removes environmental noise, identifies obstacle clusters, and extracts the obstacle positions;

[0125] S3, the main control processing module calculates the relative speed and collision time between the crane and the obstacle based on the real-time running speed and displacement fed back by the crane encoder;

[0126] S4, the main control processing module dynamically classifies the current protection level based on the comparison results between the collision time and the preset multi-level thresholds, generates the corresponding hierarchical control instructions and outputs them;

[0127] S5, the redundant safety execution module monitors the communication status and heartbeat signal of the main control processing module in real time. When it detects that the main control processing module has a communication interruption, software deadlock, or receives an emergency stop signal, it bypasses the software logic to directly cut off the power supply to the power circuit and trigger the mechanical brake, so that the system enters a fault-safe self-locking state.

[0128] This embodiment describes the entire process from environmental perception to hardware braking through five steps, S1 to S5. S1 and S2 are responsible for environmental perception and obstacle recognition, transforming the raw point cloud into structured obstacle information; S3 integrates the perceived data and its own state data to calculate a predictive collision time index; S4 makes hierarchical decisions based on the collision time value and outputs control commands that match the risk level; S5 serves as a hardware-level safety backup, remaining in standby mode when the main control module is functioning normally and only intervening in abnormal situations. This ensures both flexible control under normal operating conditions and rigid protection under abnormal operating conditions.

[0129] The various embodiments of the present invention have now been described in detail. To avoid obscuring the concept of the invention, some details known in the art have not been described. Those skilled in the art will fully understand how to implement the technical solutions disclosed herein based on the above description.

[0130] The embodiments described above only illustrate some implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A crane lidar collision avoidance system based on collision time-based graded protection, characterized in that, include: The lidar sensing module is installed on the main structure of the crane and is configured to scan the working area and acquire three-dimensional point cloud data of obstacles. The main control processing module is communicatively connected to the lidar sensing module and the crane drive mechanism. It is configured to preprocess lidar point cloud data to identify obstacles, calculate relative speed and collision time based on the crane's real-time running speed and displacement, and dynamically divide multi-level protection zones and output corresponding hierarchical control commands based on the comparison results of collision time and preset multi-level thresholds. The communication module is used for data transmission between the lidar sensing module and the main control processing module. The redundant safety execution module adopts a dual-channel safety relay architecture and is connected in parallel with the main control processing module. The input terminal of the safety relay receives the emergency stop signal from the main control processing module and the external hard-wired emergency stop signal, and the output terminal is directly connected in series in the power circuit contactor coil and brake coil circuit of the crane drive motor. The redundant safety execution module is configured to forcibly cut off the power supply to the power circuit and trigger the mechanical brake when it detects a communication interruption, software deadlock, or emergency stop signal trigger of the main control processing module, thereby achieving fault-safe self-locking.

2. The crane lidar collision avoidance system based on collision time-based graded protection according to claim 1, characterized in that, The main control processing module is equipped with a virtual fence logic unit: The virtual fence logic unit dynamically divides the work area into a safe zone, a warning zone, a deceleration zone, and an emergency stop zone based on the collision time and a preset multi-level threshold. In the safe zone, if the collision time is greater than the first threshold, the system maintains normal operation and does not output intervention commands. In the warning zone, when the collision time is greater than the second threshold and less than or equal to the first threshold, an audible and visual alarm signal is output. In the deceleration zone, when the collision time is greater than the third threshold and less than or equal to the second threshold, a power cut-off command is output, and inertial gliding is used to trigger electric braking. In the emergency stop zone, when the collision time is less than or equal to the third threshold, a dual-channel emergency stop command is output to control the combined braking of the electric brake and the mechanical brake. Wherein, the first threshold is greater than the second threshold, and the second threshold is greater than the third threshold.

3. The crane lidar collision avoidance system based on collision time-based graded protection according to claim 1, characterized in that, The main control processing module includes: The point cloud denoising unit uses a statistical filtering algorithm to calculate the average k-nearest neighbor distance for each point and remove isolated noise points that exceed the standard deviation of the global mean by a preset multiple. The point cloud clustering unit uses Euclidean distance clustering algorithm or voxel grid segmentation algorithm to segment spatially adjacent point clouds into independent obstacle clusters and remove noise clusters with fewer than a preset threshold number of points. The obstacle centroid extraction unit extracts the centroid coordinates of each obstacle cluster as the obstacle location feature.

4. The crane lidar collision avoidance system based on collision time-based graded protection according to claim 3, characterized in that: Before the point cloud denoising unit uses a statistical filtering algorithm to remove isolated noise points, the main control processing module first performs voxel grid downsampling on the point cloud data. In the voxel grid downsampling, the three-dimensional working space is divided into several voxel grids with preset side lengths. The centroid of all point clouds in a single voxel grid is taken as the representative point of that voxel grid to complete the point cloud compression.

5. The crane lidar collision avoidance system based on collision time-based graded protection according to claim 3, characterized in that, In the Euclidean distance clustering algorithm, point clouds with spatial spacing less than a preset clustering distance threshold are grouped into the same obstacle cluster, and obstacle clusters with a number of point clouds less than a preset number of points are filtered out.

6. The crane lidar collision avoidance system based on collision time-based graded protection according to claim 3, characterized in that, The main control processing module includes: The relative distance calculation unit calculates the relative distance between the crane and the obstacle based on the obstacle's center coordinates and the real-time displacement fed back by the crane's encoder. The relative velocity calculation unit collects a series of relative distance sequences and corresponding timestamps from multiple past frames, and obtains the relative velocity by fitting the relationship between relative distance and time through linear regression. The collision time calculation unit calculates the collision time by taking the ratio of the current relative distance to the absolute value of the relative speed when the relative speed is negative, i.e., when the obstacle is approaching the crane; when the relative speed is greater than or equal to zero, it is determined that there is no risk of collision.

7. The crane lidar collision avoidance system based on collision time-based graded protection according to claim 1, characterized in that, The dual-channel safety relay architecture of the redundant safety execution module is configured as follows: During normal operation, cross-validation is performed on both channels; When any channel experiences contact adhesion, wire breakage, or inconsistent input signals, the monitoring circuit immediately cuts off the power supply to all output coils and forcibly disconnects the main contacts using a spring reset mechanism. When the main control processing module fails to send a heartbeat maintenance signal within the specified period, the safety relay enters a de-energized state and forcibly disconnects the output.

8. The crane lidar collision avoidance system based on collision time-based graded protection according to claim 1, characterized in that: The lidar sensing module includes multiple lidars, which are installed at different structural positions on the crane. The main control processing module is configured to stitch together the point cloud data from the multiple lidars to generate a full-view point cloud data.

9. The crane lidar collision avoidance system based on collision time-based graded protection according to claim 1, characterized in that: The lidar sensing module includes at least one single-line lidar or a multi-line lidar. The scanning frequency of the lidar sensing module is greater than or equal to 10Hz, and the detection distance range is from 0.1m to 50m. The lidar sensing module is fixedly installed on the crane's main trolley or auxiliary trolley traveling mechanism and moves synchronously with the crane.

10. A crane lidar collision avoidance method based on collision time-based graded protection, employing the crane lidar collision avoidance system as described in any one of claims 1 to 9, characterized in that, Includes the following steps: The lidar sensing module periodically scans and acquires 3D point cloud data of the work area, which is then uploaded to the main control processing module via the communication module. The main control processing module preprocesses the point cloud data, removes environmental noise, identifies obstacle clusters, and extracts obstacle locations; The main control processing module calculates the relative speed and collision time between the crane and the obstacle based on the real-time operating speed and displacement fed back by the crane encoder. The main control processing module dynamically classifies the current protection level based on the comparison results between the collision time and the preset multi-level thresholds, generates corresponding hierarchical control commands, and outputs them. The redundant safety execution module monitors the communication status and heartbeat signal of the main control processing module in real time. When it detects that the main control processing module has a communication interruption, software deadlock, or receives an emergency stop signal, it bypasses the software logic to directly cut off the power supply to the power circuit and trigger the mechanical brake, so that the system enters a fault-safe self-locking state.

Images