Loading and transporting machinery safety intelligent auxiliary system and method

Abstract

The invention discloses a safe and intelligent auxiliary system and method for loading and transporting machinery, relates to the technical field of intelligent loading and unloading, and is suitable for efficient cargo loading in a narrow space, such as a warehouse and an underground parking lot. Rapid adjustment can be achieved under the conditions of different carriage heights, sizes and shapes, and various loading requirements are met; rolling, lifting and folding functions are integrated, and the loading and unloading efficiency is remarkably improved; a laser radar and a camera are used for generating a real-time 3D map, obstacles, stand columns and truck parking positions are dynamically recognized, and in truck butt joint point positioning, optical marks and multi-sensor data are combined, so that butt joint is more accurate; a reinforcement learning path planning algorithm is introduced, a path is dynamically optimized according to real-time environment data, so that the platform can efficiently reach a target point, the obstacle avoidance performance is improved, the advancing speed and direction are dynamically adjusted based on the obstacle distance, and safe operation is ensured; in the butt joint stage, the relative position of the telescopic and foldable loading platform is finely adjusted, and accurate butt joint with the tail of the truck is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent loading and unloading, and in particular to a safe intelligent auxiliary system and method for loading and unloading machinery. Background Art

[0002] Current loading equipment has low loading efficiency in crowded warehouses or buildings, and in some cases even requires manual assistance to complete loading. This is mainly because there are many load-bearing columns or partition columns in the building, and the goods are densely stacked. Vehicles are restricted by the small space when loading. In this scenario, whether using existing loading and unloading tools or other methods, it is difficult to complete the loading task quickly and efficiently.

[0003] In addition, current shipping equipment has obvious deficiencies in balance performance. During the loading process, if heavy cargo is concentrated or unevenly loaded, it is easy to cause the platform to tilt or even the cargo to slip, which not only endangers the safety of the equipment and cargo, but may also cause operations to be interrupted.

[0004] To address these problems, some loading equipment sets load limits and operators plan the cargo placement in advance. However, this method is difficult to adjust in real time according to actual loading conditions. At the same time, due to the lack of optimization of the cargo loading sequence, large or heavy cargo is placed in the carriage more randomly, resulting in the shift of the center of gravity of the carriage, which not only affects the overall loading stability and transportation safety, but also reduces the loading efficiency. It is difficult to effectively improve the utilization of the carriage space. Therefore, a loading machinery safety intelligent assistance system and method are urgently needed to solve such problems. Summary of the invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a loading machinery safety intelligent assistance system and method to solve the problem that in a crowded warehouse, due to various reasons such as surrounding building columns and densely stacked goods, during the process of loading goods into the vehicle entering the warehouse, due to the limitation of small space, no matter which loading and unloading tools are currently used, it is impossible to load goods quickly and effectively.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] On the one hand, the present invention provides a safe intelligent auxiliary system for loading and unloading machinery, which includes a mobile platform and an auxiliary control center:

[0009] The mobile platform includes: a retractable loading platform, a retractable and foldable loading platform and a load-bearing walking mechanism;

[0010] The auxiliary control center includes: environmental perception module, path planning module, dynamic load balancing module and multi-task collaborative scheduling module;

[0011] The retractable loading platform can be adjusted in height electrically or manually, and the cargo can be moved electrically or manually;

[0012] The retractable and foldable loading platform is electrically or manually adjustable in height, the relative position of the retractable loading platform is electrically or manually adjustable, and the carried goods are electrically or manually moved;

[0013] Environmental perception module, which is used to perceive the work site environment in real time and generate dynamic 3D work maps to identify trucks, obstacles and loading and unloading area information;

[0014] The path planning module is responsible for setting and executing the movement path of the mobile platform from the initial position to the truck docking point;

[0015] Dynamic loading balance module, which controls the balance of the mobile platform during loading, monitors and adjusts the load distribution of the cargo;

[0016] Multi-task collaborative scheduling module coordinates the allocation of work tasks between forklifts and platforms.

[0017] Furthermore, the retractable and foldable loading platform integrates a forklift function, which can lift the cargo at the end for loading operations, and the retractable and foldable loading platform integrates a load pressure sensor network and a hydraulic height adjustment unit;

[0018] A network of load pressure sensors is distributed on the surface of the retractable and foldable loading platform to monitor the weight distribution, volume and location of the cargo in real time;

[0019] A hydraulic height adjustment unit adjusts the height of each support point of the platform based on the output of the dynamic load balancing module;

[0020] The load-bearing walking mechanism in the mobile platform supports the retractable loading platform and the retractable and foldable loading platform. It has powered rollers, can move in four horizontal directions, and has a locking device and hydraulic height adjustment function.

[0021] Furthermore, the working states of the mobile platform include a standby state, an unfolded state, a docked state, and a loaded state;

[0022] Standby state: When the platform is not working, the retractable loading platform and the retractable and foldable loading platform maintain an L-shaped vertical state. After the truck stops in the parking area, it starts loading operations in the loading area and enters the unfolded state;

[0023] Unfolded state: When the truck enters the loading area, the retractable loading platform and the retractable and foldable loading platform are unfolded to a T-shaped state, and the height of the retractable and foldable loading platform is adjusted to be consistent with the height of the truck compartment, and enters the docking state;

[0024] Docking state: the mobile platform moves as a whole to the rear of the carriage, the retractable and foldable loading platform docks with the carriage, the retractable and foldable loading platform adjusts its position to go deep into the carriage, and enters the loading state;

[0025] Loading status: The retractable and foldable loading platform goes further into the carriage and reserves enough space for the goods. The goods are delivered to the retractable loading platform by a forklift, rolled to the retractable and foldable loading platform, and finally lifted into the carriage by the forklift function of the retractable and foldable loading platform. The mobile platform keeps moving backwards before the carriage is full of goods to make room for new goods.

[0026] Furthermore, the environment perception module includes:

[0027] LiDAR and cameras are used to generate a three-dimensional map of the site and identify obstacles in the warehouse environment;

[0028] The truck positioning unit, combined with the optical marker or RFID tag at the rear of the truck, locates the specific position and docking point of the truck;

[0029] The path planning module uses a reinforcement learning path planning algorithm to calculate the moving path of the mobile platform based on the 3D map of the work site. During the path execution process, the navigation path is adjusted according to the data of the environmental perception module to avoid obstacles;

[0030] It is also responsible for docking the retractable and foldable loading platform to the rear of the truck.

[0031] Further, a dynamic load balancing module calculates the load distribution of the platform based on the cargo weight distribution monitored by the load pressure sensor network and outputs the height of the dynamic hydraulic height adjustment unit;

[0032] The multi-task collaborative scheduling module determines the loading order of goods (such as loading heavy goods to the bottom of the carriage first) based on the weight, volume and location of the goods obtained by the load pressure sensor network; and assigns tasks to forklifts and platforms so that there is no conflict between the two operations.

[0033] In another aspect, the present invention provides a method for intelligently assisting safe loading and unloading machinery, comprising:

[0034] Step S1, environmental perception,

[0035] Generate a real-time 3D map of the warehouse site based on LiDAR and cameras to identify obstacles in the loading and unloading area, the location of pillars, and the docking position of trucks; use optical markers at the rear of the carriage to locate the docking point of the truck, obtain a complete work site model and the location of the truck docking point;

[0036] Step S2: dynamic path planning and initial positioning.

[0037] Based on the 3D map in step S1, combined with the reinforcement learning path planning algorithm, a path is generated for the mobile platform from the current position to the truck docking point;

[0038] In step S2, the mobile platform avoids obstacles along the way and adaptively adjusts the travel speed and direction. After reaching the target point, the relative position of the retractable and foldable loading platform and the carriage is adjusted so that the retractable and foldable loading platform docks with the rear of the carriage and enters the loading state.

[0039] Step S3, dynamic load balancing optimization,

[0040] Before loading, the dynamic load balancing module uses a balance optimization algorithm to calculate the cargo placement based on the cargo weight distribution monitored in real time by the load pressure sensor network, and then the hydraulic height adjustment unit adjusts the height to avoid center of gravity shift;

[0041] Step S4, multi-task collaborative loading,

[0042] The multi-task collaborative scheduling module generates an optimized loading queue based on the cargo list and the spatial layout of the trucks:

[0043] Prioritize loading heavy cargo to the bottom of the carriage or near the front of the carriage, and place larger cargo on both sides of the carriage;

[0044] The mobile platform receives the cargo from the forklift and transfers the cargo to the retractable and foldable loading platform. The forklift function is used to load the cargo into the carriage one by one. When the cargo approaches the rear of the carriage, the mobile platform automatically moves back to release new loading space.

[0045] Step S5: platform reset.

[0046] After the loading task is completed, the auxiliary control center automatically returns to the initial position according to the path planning result of step S2.

[0047] Furthermore, in step S1, the method of generating a 3D map of the warehouse site is as follows:

[0048] Assume that the environmental point cloud data acquired by the laser radar is P L , P L ∈R N×3 , where N is the number of points in the point cloud, each point contains three-dimensional coordinates (x, y, z), R N×3 represents a matrix with N rows and 3 columns, each element is a real number, R represents the set of real numbers, P L ={(x i ,y i , z i )|i=1,2,...,N},where x i ,y i , zi Represents the three-dimensional space coordinates of point i in the point cloud;

[0049] The point cloud data is segmented on the ground, and the least square method is used to fit the plane equation: z = ax + by + c, where a, b, c represent the plane parameters obtained by the least square fitting of the point cloud data, z represents the height of each point in the point cloud, and the ground points satisfy the fitted plane model;

[0050] Calculate the distance from the point to the plane and remove points whose distance is less than the threshold ∈ g The ground points are obtained to obtain the non-ground point cloud P′ L :

[0051] P′ L ={(x i ,y i , z i )|d i >∈ g}, where d i Represents the vertical distance from point i to the fitting plane, ∈ g Indicates the ground distance threshold;

[0052] In step S1, depth information is generated based on the camera image:

[0053] The camera captures the scene image I C , and use structured light or binocular vision to calculate the depth map:

[0054] Among them, D C (x, y) represents the depth value corresponding to the pixel point (x, y), f represents the focal length of the camera, B represents the baseline length of the binocular camera, d L , d r Indicates the disparity value of the corresponding pixel points of the left and right cameras;

[0055] Convert the depth map to a point cloud P C :

[0056] Among them, c x , c y Indicates the camera focus coordinates, f x , f y Represents the components of the camera focal length in the x and y directions;

[0057] The point cloud is fused with the image data, and the camera point cloud P is transformed using the external parameter matrix T, including the rotation matrix R and the translation vector t. C Convert to the laser radar coordinate system to obtain the fused point cloud P F :

[0058] PF =R·P C +t,P F =P′ L UP F , where R represents the rotation matrix from the camera coordinate system to the lidar coordinate system, and t represents the translation vector;

[0059] For the fused point cloud P F Perform clustering and use the DBSCAN algorithm to separate obstacles, pillars, and other objects;

[0060] In step S1, the truck docking point is located:

[0061] Use the optical markers at the rear of the vehicle to locate the docking point, combined with the camera image I C Calculate the marker position:

[0062] (x m ,y m , z m )=R·(x p ,y p , z p )+t, where (x p ,y p , z p ) represents the three-dimensional coordinates of the optical marker in the camera coordinate system, (x m ,y m , z m ) represents the coordinates of the optical marker in the laser radar coordinate system;

[0063] According to the marking position (x m ,y m , z m ) Calculate the position of the truck docking point relative to the mobile platform:

[0064] Δ x =x m -x p , Δ y =y m -y p , Δ z =z m -z p , where Δ x , Δ y and Δ z Indicates the offset of the mobile platform relative to the truck docking point, (x p ,y p , z p ) represents the current position of the mobile platform.

[0065] Furthermore, in step S2, the path is generated based on the 3D map in the following manner:

[0066] Define the state of the mobile platform as s t =(x t ,y t ,θ t ), indicating the position at time t (x t ,y t ) and the orientation θ t ,

[0067] Define the action space as a t =(v t ,ω t ), where v t represents the linear velocity, ω t represents the angular velocity,

[0068] Define the reward function R(s t , a t )for:

[0069] R(s t , a t )=-α·d g -β·d o -γ·|ω t |, where d g Represents the Euclidean distance between the current position and the target point, d o It represents the distance between the current position and the nearest obstacle. α, β, and γ represent weight parameters, which balance the requirements of reaching the target point, avoiding obstacles, and smoothing the path.

[0070] Select the deep reinforcement learning algorithm DQN and DDPG update strategy π(a t |s t ), optimize the expected cumulative reward: Where γ represents the discount factor, which controls the weight of future rewards, J(π) represents the expected cumulative reward of strategy π, E[·] is the expectation operator, and T represents the time step termination condition;

[0071] In step S2, the path is adjusted in real time:

[0072] The target path generated by path planning is expressed as That is, the target path generated by path planning is composed of N discrete points, and the two-dimensional plane coordinates of each point are (x i ,y i ), the mobile platform adjusts the speed and direction to track the path in real time, the linear velocity v t and angular velocity ω t for:

[0073] v t =K v ·d t ,ωt =K ω ·(θ g -θ t ),in, Indicates the distance from the current position of the platform to the target point. Indicates the direction of the current target point, K v , K ω represents the control gain coefficient, (x t ,y t ) represents the current position of the mobile platform, (x g ,y g ) represents the target point coordinates;

[0074] If an obstacle is detected, the distance d from the obstacle to the platform is calculated o :

[0075] If d o <∈ s , adjust the target direction:

[0076] θ t =θ t +Δθ,Δθ=sgn(ω t )·K a ·(∈ s -d o ), where ∈ s Represents the safety distance threshold, K a represents the obstacle avoidance adjustment coefficient, (x o ,y o ) represents the position coordinates of the obstacle, P F is the fused point cloud data set;

[0077] In step S2, the platform is docked with the rear of the carriage:

[0078] When the platform reaches the target point, it enters the fine-tuning stage, adjusting the position of the retractable and foldable loading platform relative to the rear of the carriage and fine-tuning the offset:

[0079] Δ x =x g -x t , Δ y =y g -y t , Δ z =z g -z t , where x g ,y g , z g Indicates the coordinates of the docking point at the rear of the carriage, x t ,y t , z tIndicates the current position of the retractable and foldable loading platform;

[0080] Adjust the platform height h using the hydraulic unit t :h t =h g +K h ·(Δ z ), where h g Indicates the height of the docking point at the rear of the carriage, K h Indicates the height adjustment factor.

[0081] Furthermore, in step S3, balance optimization is performed based on the real-time weight distribution:

[0082] Assume that the sensor network of the loading platform consists of M×N distribution points, and the sensor is located at position (x i，j , Y i，j ) is recorded as f i，j , the total pressure matrix is ​​F: F = {f i，j |1≤i≤M,1≤j≤N};

[0083] Calculate the weight distribution of the sensor sampling points: Where g is the acceleration due to gravity, w is the i，j is the cargo weight at point (i, j);

[0084] Calculate the total weight W and the center of gravity of the platform (x c ,y c ):

[0085]

[0086] In step S3, a balanced optimization algorithm is used to calculate the placement position:

[0087] The goal is to make the center of gravity of the platform (x c ,y c ) approaches the target equilibrium point (x t ,y t ), objective function for:

[0088]

[0089] Optimize the placement of goods, assuming the weight of the new goods is w n , its candidate placement point is (x n ,y n ), optimize the model for:

[0090] Enumerate candidate points (x n ,y n ) or gradient descent method to solve the optimal point;

[0091] In step S3, the hydraulic height is adjusted in the following manner:

[0092] The adjustment target is defined as: Keep the platform surface balanced so that the height of the hydraulic support point is h k，l In line with the optimized inclination requirements, the platform hydraulic support point is set as P k，l , whose coordinates are (x k,l ,y k,l ), the adjusted height h k，l for:

[0093] h k，l =h0+K·(x k,l -x c )+K·(y k，l -y c ), where h0 represents the initial height of the platform, and K represents the adjustment coefficient, which is determined based on the calculation results of the dynamic loading balance module;

[0094] According to the height difference Δh of the hydraulic support point k，l Adjust the final inclination angle of the platform:

[0095] Among them, θ x ,θ y Indicates the inclination angle of the platform in the x and y directions.

[0096] Furthermore, in step S4, the loading queue is adjusted in the following manner:

[0097] Assume that the goods collection is g per piece of goods i The properties include: weight m i , volume v i , the initial position coordinates of the goods p i =(x i ,y i , z i ) and loading priority d i ;

[0098] Define loading priority according to carriage layout i : Where M represents the total weight of the cargo collection, V represents the total volume of the cargo collection, C represents the reference coordinates of the truck loading point, and dist(p i , C) represents the distance between the initial position coordinates and the reference coordinates, α, β, γ represent the priority weight coefficients, and heavy cargo m i and bulky cargo i The higher the weight, the higher the priority of goods closer to the loading point;

[0099] In step S4, an optimized loading queue is generated:

[0100] Divide the carriage into the front, middle, rear, and left and right sides, using the partition index L j，k express:

[0101] L j，k ={(x, y, z)|x∈[x j,min , x j,max ],y∈[y k,min ,y k,max ]}, where j represents the longitudinal area index of the carriage, including the front, middle and rear, k represents the transverse area index of the carriage, including the left and right, and x j,min , x j，max and k,min ,y k,max Represents the minimum and maximum boundaries of the partition in the x and y directions respectively;

[0102] The objective function of cargo distribution is defined as

[0103] Among them, M j，k Indicates area L j，k Current accumulated weight, V j，k Indicates area L j，k The current cumulative volume, ω1, ω2, ω3 represent the regional allocation weight coefficients, dist(p i , L j，k ) indicates the cargo g i Current location p i =(x i ,y i , z i ) to compartment L j，k distance;

[0104] g per piece of goods i Assign to target area

[0105] By priority d i and regional allocation Sorting, generating optimized queue Q:

[0106] In step S4, loading is performed:

[0107] Mobile platform receives cargo from forklift i , update its coordinates to the current position of the platform Transfer to the retractable and foldable loading platform: in, represents the coordinates after transfer, T represents the coordinate transformation matrix from the mobile platform to the loading platform;

[0108] The retractable and foldable loading platform uses the forklift function to load the goods into the target area Adjust cargo position to optimize distribution: Among them, Δp i Indicates the offset between the current position of the cargo and the center of the target area; adjust the cargo position: in, is the final cargo location;

[0109] When the cargo is loaded to the rear and there is insufficient space in the carriage, the mobile platform moves backward.

[0110] The beneficial effects of the present invention are:

[0111] The invention is suitable for efficient cargo loading in narrow spaces, such as warehouses and underground parking lots; it can be quickly adjusted to different compartment heights, sizes and shapes to meet a variety of loading requirements; it integrates rolling, lifting and folding functions to significantly improve loading and unloading efficiency;

[0112] The present invention utilizes the fusion of laser radar and camera to generate real-time 3D maps, dynamically identify obstacles, pillars and truck parking positions, improve the accuracy and real-time performance of environmental perception, adapt to dynamic scene changes, and combine optical markers and multi-sensor data in the positioning of truck docking points to make docking more accurate.

[0113] The present invention introduces a reinforcement learning path planning algorithm, dynamically optimizes the path according to real-time environmental data, enables the platform to efficiently reach the target point, improves obstacle avoidance performance, and dynamically adjusts the travel speed and direction based on the obstacle distance to ensure safe operation; during the docking stage, fine-tunes the relative position of the retractable and foldable loading platform to ensure accurate docking with the rear of the truck.

[0114] The present invention monitors the weight distribution of goods in real time, calculates the best placement position in combination with a balance optimization algorithm to avoid center of gravity shift, and a hydraulic height adjustment unit dynamically adjusts the height of the platform support point to ensure that the loading platform remains balanced under any loading conditions, thereby improving loading safety and uniformity of cargo distribution.

[0115] The present invention generates an optimized loading queue according to the weight, volume and initial position of the cargo, prioritizes loading heavy cargo to the bottom of the compartment or near the front of the compartment, reasonably distributes large cargo to both sides of the compartment, and coordinates the mobile platform with the forklift to avoid task conflicts, dynamically retreats to release loading space, and improves the overall loading efficiency; the sorting logic of the loading queue makes the cargo inside the truck more stably arranged, which is convenient for subsequent transportation. BRIEF DESCRIPTION OF THE DRAWINGS

[0116] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0117] Figure 1 It is a structural schematic diagram of the intelligent safety assistance system for loading machinery of the present invention;

[0118] Figure 2 It is a schematic diagram of the process flow of the intelligent auxiliary method for safe loading machinery of the present invention;

[0119] Figure 3 It is a structural schematic diagram of the mobile platform of the present invention in a standby state;

[0120] Figure 4 This is a structural schematic diagram of the mobile platform of the present invention in an unfolded state.

[0121] Figure 5 It is a structural schematic diagram of the mobile platform of the present invention in a docking state;

[0122] Figure 6 It is a structural schematic diagram of the mobile platform of the present invention in the loading state. DETAILED DESCRIPTION

[0123] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0124] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0125] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0126] Example 1, reference Figure 1 This embodiment provides a safe intelligent auxiliary system for loading and unloading machinery, including: a mobile platform and an auxiliary control center,

[0127] The mobile platform includes: a retractable loading platform, a retractable and foldable loading platform and a load-bearing walking mechanism;

[0128] The auxiliary control center includes: environmental perception module, path planning module, dynamic load balancing module and multi-task collaborative scheduling module;

[0129] The retractable loading platform can be adjusted in height electrically or manually, and the cargo can be moved electrically or manually;

[0130] The retractable and foldable loading platform is electrically or manually adjustable in height, the relative position of the retractable loading platform is electrically or manually adjustable, and the carried goods are electrically or manually moved;

[0131] The retractable and foldable loading platform integrates forklift function, which can lift the goods at the end for loading operation. The retractable and foldable loading platform integrates load pressure sensor network and hydraulic height adjustment unit;

[0132] A network of load pressure sensors is distributed on the surface of the retractable and foldable loading platform to monitor the weight distribution, volume and location of the cargo in real time;

[0133] A hydraulic height adjustment unit adjusts the height of each support point of the platform based on the output of the dynamic load balancing module;

[0134] The load-bearing walking mechanism in the mobile platform supports the retractable loading platform and the retractable and foldable loading platform, has powered rollers, can move in four horizontal directions, and has a locking device and hydraulic height adjustment function;

[0135] The working states of the mobile platform include standby state, deployment state, docking state and loading state;

[0136] Standby state: When the platform is not working, the retractable loading platform and the retractable and foldable loading platform maintain an L-shaped vertical state. After the truck stops in the parking area, it starts loading operations in the loading area and enters the unfolded state;

[0137] Unfolded state: When the truck enters the loading area, the retractable loading platform and the retractable and foldable loading platform are unfolded to a T-shaped state, and the height of the retractable and foldable loading platform is adjusted to be consistent with the height of the truck compartment, and enters the docking state;

[0138] Docking state: the mobile platform moves as a whole to the rear of the carriage, the retractable and foldable loading platform docks with the carriage, the retractable and foldable loading platform adjusts its position to go deep into the carriage, and enters the loading state;

[0139] Loading status: The retractable and foldable loading platform goes deeper into the carriage and reserves enough space for the goods. The goods are delivered to the retractable loading platform by a forklift, rolled to the retractable and foldable loading platform, and finally lifted into the carriage by the forklift function of the retractable and foldable loading platform. The mobile platform keeps moving backwards before the carriage is full of goods to make room for new goods.

[0140] Environmental perception module, which is used to perceive the work site environment in real time and generate dynamic 3D work maps to identify trucks, obstacles and loading and unloading area information;

[0141] The environmental perception module includes:

[0142] LiDAR and cameras are used to generate a three-dimensional map of the site and identify obstacles in the warehouse environment;

[0143] The truck positioning unit, combined with the optical marker or RFID tag at the rear of the truck, locates the specific position and docking point of the truck;

[0144] The path planning module is responsible for setting and executing the movement path of the mobile platform from the initial position to the truck docking point;

[0145] The path planning module uses a reinforcement learning path planning algorithm to calculate the moving path of the mobile platform based on the 3D map of the work site. During the path execution process, the navigation path is adjusted according to the data of the environmental perception module to avoid obstacles;

[0146] Also responsible for docking the retractable and foldable loading platform to the rear of the truck;

[0147] Dynamic loading balance module, which controls the balance of the mobile platform during loading, monitors and adjusts the load distribution of the cargo;

[0148] Dynamic load balancing module, which calculates the load distribution of the platform based on the cargo weight distribution monitored by the load pressure sensor network and outputs the height of the dynamic hydraulic height adjustment unit;

[0149] Multi-task collaborative scheduling module coordinates the task allocation between forklifts and platforms;

[0150] The multi-task collaborative scheduling module determines the loading order of goods (such as loading heavy goods to the bottom of the carriage first) based on the weight, volume and location of the goods obtained by the load pressure sensor network; and assigns tasks to forklifts and platforms so that there is no conflict between the two operations.

[0151] Example 2, reference Figure 2 This embodiment provides a method for intelligent assistance of shipping machinery safety, comprising the following steps:

[0152] Step S1, environmental perception,

[0153] Generate a real-time 3D map of the warehouse site based on LiDAR and cameras to identify obstacles in the loading and unloading area, the location of pillars, and the docking position of trucks; use optical markers at the rear of the carriage to locate the docking point of the truck, obtain a complete work site model and the location of the truck docking point;

[0154] In step S1, the method of generating a 3D map of the warehouse site is as follows:

[0155] Assume that the environmental point cloud data acquired by the laser radar is P L , P L ∈R N×3 , where N is the number of points in the point cloud, each point contains three-dimensional coordinates (x, y, z), R N×3 represents a matrix with N rows and 3 columns, each element is a real number, R represents the set of real numbers, P L ={(x i ,y i , z i )|i=1,2,...,N},where x i ,y i , z i Represents the three-dimensional space coordinates of point i in the point cloud;

[0156] The point cloud data is segmented on the ground, and the least square method is used to fit the plane equation: z = ax + by + c, where a, b, c represent the plane parameters obtained by the least square fitting of the point cloud data, z represents the height of each point in the point cloud, and the ground points satisfy the fitted plane model;

[0157] Calculate the distance from the point to the plane and remove points whose distance is less than the threshold ∈ g The ground points are obtained to obtain the non-ground point cloud P′ L :

[0158] Among them, d i Represents the vertical distance from point i to the fitting plane, ∈ g Indicates the ground distance threshold;

[0159] In step S1, depth information is generated based on the camera image:

[0160] The camera captures the scene image I C , and use structured light or binocular vision to calculate the depth map:

[0161] Among them, D C (x, y) represents the depth value corresponding to the pixel point (x, y), f represents the focal length of the camera, B represents the baseline length of the binocular camera, d L , d r Indicates the disparity value of the corresponding pixel points of the left and right cameras;

[0162] Convert the depth map to a point cloud P C :

[0163] Among them, c x , c y Indicates the camera focus coordinates, f x , f y Represents the components of the camera focal length in the x and y directions;

[0164] The point cloud is fused with the image data, and the camera point cloud P is transformed using the external parameter matrix T, including the rotation matrix R and the translation vector t. C Convert to the laser radar coordinate system to obtain the fused point cloud P F :

[0165] P F =R·P C +t,P F =P′ L UP F , where R represents the rotation matrix from the camera coordinate system to the lidar coordinate system, and t represents the translation vector;

[0166] For the fused point cloud P F Perform clustering and use the DBSCAN algorithm to separate obstacles, pillars, and other objects;

[0167] In step S1, the truck docking point is located:

[0168] Use the optical markers at the rear of the vehicle to locate the docking point, combined with the camera image I C Calculate the marker position:

[0169] (x m ,y m , z m )=R·(x p ,y p , z p )+t, where (x p ,y p , z p ) represents the three-dimensional coordinates of the optical marker in the camera coordinate system, (x m ,y m , z m ) represents the coordinates of the optical marker in the laser radar coordinate system;

[0170] According to the marking position (x m ,y m , z m ) Calculate the position of the truck docking point relative to the mobile platform:

[0171] Δ x =xm -x p , Δ y =y m -y p , Δ z =z m -z p , where Δ x , Δ y and Δ z Indicates the offset of the mobile platform relative to the truck docking point, (x p ,y p , z p ) represents the current position of the mobile platform;

[0172] Specifically, through step S1, a complete real-time 3D map of the warehouse is generated and the truck docking point is located, and key data P is output. F (fused point cloud) and (Δ x , Δ y , Δ z )(Door point offset).

[0173] Step S2: dynamic path planning and initial positioning.

[0174] Based on the 3D map in step S1, combined with the reinforcement learning path planning algorithm, a path is generated for the mobile platform from the current position to the truck docking point;

[0175] In step S2, the mobile platform avoids obstacles along the way and adaptively adjusts the travel speed and direction. After reaching the target point, the relative position of the retractable and foldable loading platform and the carriage is adjusted so that the retractable and foldable loading platform docks with the rear of the carriage and enters the loading state.

[0176] In step S2, the path is generated based on the 3D map in the following manner:

[0177] Define the state of the mobile platform as s t =(x t ,y t ,θ t ), indicating the position at time t (x t ,y t ) and the orientation θ t ,

[0178] Define the action space as a t =(v t ,ω t ), where v t represents the linear velocity, ω t represents the angular velocity,

[0179] Define the reward function R(s t , a t)for:

[0180] R(s t , a t )=-α·d g -β·d o -γ·|ω t |, where d g Represents the Euclidean distance between the current position and the target point, d o It represents the distance between the current position and the nearest obstacle. α, β, and γ represent weight parameters, which balance the requirements of reaching the target point, avoiding obstacles, and smoothing the path.

[0181] Select the deep reinforcement learning algorithm DQN and DDPG update strategy π(a t |s t ), optimize the expected cumulative reward: Where γ represents the discount factor, which controls the weight of future rewards, J(π) represents the expected cumulative reward of strategy π, E[·] is the expectation operator, and T represents the time step termination condition;

[0182] In step S2, the path is adjusted in real time:

[0183] The target path generated by path planning is expressed as That is, the target path generated by path planning is composed of N discrete points, and the two-dimensional plane coordinates of each point are (x i ,y i ), the mobile platform adjusts the speed and direction to track the path in real time, the linear velocity v t and angular velocity ω t for:

[0184] v t =K v ·d t ,ω t =K ω ·(θ g -θ t ),in, Indicates the distance from the current position of the platform to the target point. Indicates the direction of the current target point, K v , K ω represents the control gain coefficient, (x t ,y t ) represents the current position of the mobile platform, (x g ,y g ) represents the target point coordinates;

[0185] If an obstacle is detected, the distance d from the obstacle to the platform is calculated o :

[0186] If d o <∈ s , adjust the target direction:

[0187] θ t =θ t +Δθ,Δθ=sgn(ω t )·K a ·(∈ s -d o ), where ∈ s Represents the safety distance threshold, K a represents the obstacle avoidance adjustment coefficient, (x o ,y o ) represents the position coordinates of the obstacle, P F is the fused point cloud data set;

[0188] In step S2, the platform is docked with the rear of the carriage:

[0189] When the platform reaches the target point, it enters the fine-tuning stage, adjusting the position of the retractable and foldable loading platform relative to the rear of the carriage and fine-tuning the offset:

[0190] Δ x =x g -x t , Δ y =y g -y t , Δ z =z g -z t , where x g ,y g , z g Indicates the coordinates of the docking point at the rear of the carriage, x t ,y t , z t Indicates the current position of the retractable and foldable loading platform;

[0191] Adjust the platform height h using the hydraulic unit t :h t =h g +K h ·(Δ z ), where h g Indicates the height of the docking point at the rear of the carriage, K h Indicates the height adjustment factor;

[0192] Specifically, through step S2, the mobile platform avoids obstacles, adaptively adjusts the path, speed and direction, and finally docks accurately at the rear of the carriage;

[0193] Step S3, dynamic load balancing optimization,

[0194] Before loading, the dynamic load balancing module uses a balance optimization algorithm to calculate the cargo placement based on the cargo weight distribution monitored in real time by the load pressure sensor network, and then the hydraulic height adjustment unit adjusts the height to avoid center of gravity shift;

[0195] In step S3, balance optimization is performed based on the real-time weight distribution:

[0196] Assume that the sensor network of the loading platform consists of M×N distribution points, and the sensor is located at position (x i，j ,y i，j ) is recorded as f i，j , the total pressure matrix is ​​F: F = {f i，j |1≤i≤M,1≤j≤N};

[0197] Calculate the weight distribution of the sensor sampling points: Where g is the acceleration due to gravity, w is the i，j is the cargo weight at point (i, j);

[0198] Calculate the total weight W and the center of gravity of the platform (x c ,y c ):

[0199]

[0200] In step S3, a balanced optimization algorithm is used to calculate the placement position:

[0201] The goal is to make the center of gravity of the platform (x c ,y c ) approaches the target equilibrium point (x t ,y t ), objective function for:

[0202]

[0203] Optimize the placement of goods, assuming the weight of the new goods is W n , its candidate placement point is (x n ,y n ), optimize the model for:

[0204] Enumerate candidate points (x n ,y n ) or gradient descent method to solve the optimal point;

[0205] In step S3, the hydraulic height is adjusted in the following manner:

[0206] The adjustment target is defined as: Keep the platform surface balanced so that the height of the hydraulic support point is h k，lIn line with the optimized inclination requirements, the platform hydraulic support point is set as P k，l , whose coordinates are (x k,l ,y k,l ), the adjusted height h k，l for:

[0207] h k，l =h0+K·(x k,l -x c )+K·(y k，L -y c ), where h0 represents the initial height of the platform, and K represents the adjustment coefficient, which is determined based on the calculation results of the dynamic loading balance module;

[0208] According to the height difference Δh of the hydraulic support point k，l Adjust the final inclination angle of the platform:

[0209] Among them, θ x ,θ y Indicates the inclination angle of the platform in the x and y directions;

[0210] Specifically, in step S3, the weight distribution of the cargo is monitored in real time, the cargo placement is optimized, the center of gravity of the platform is brought close to the target balance point, and the height of the hydraulic support point is adjusted to keep the platform stable and achieve precise dynamic loading balance;

[0211] Step S4, multi-task collaborative loading,

[0212] The multi-task collaborative scheduling module generates an optimized loading queue based on the cargo list and the spatial layout of the trucks:

[0213] Prioritize loading heavy cargo to the bottom of the carriage or near the front of the carriage, and place larger cargo on both sides of the carriage;

[0214] The mobile platform receives the cargo from the forklift and transfers the cargo to the retractable and foldable loading platform. The forklift function is used to load the cargo into the carriage one by one. When the cargo approaches the rear of the carriage, the mobile platform automatically moves back to release new loading space.

[0215] In step S4, the loading queue is adjusted in the following manner:

[0216] Assume that the goods collection is g per piece of goods i The properties include: weight m i , volume v i , the initial position coordinates of the goods p i =(x i ,y i , z i ) and loading priority d i ;

[0217] Define loading priority according to carriage layout i : Where M represents the total weight of the cargo collection, V represents the total volume of the cargo collection, C represents the reference coordinates of the truck loading point, and dist(p i , C) represents the distance between the initial position coordinates and the reference coordinates, α, β, γ represent the priority weight coefficients, and heavy cargo m i and bulky cargo i The higher the weight, the higher the priority of goods closer to the loading point;

[0218] In step S4, an optimized loading queue is generated:

[0219] Divide the carriage into the front, middle, rear, and left and right sides, using the partition index L j，k express:

[0220] L j，k ={(x, y, z)|x∈[x j,min , x j，max ],y∈[y k,min ,y k,max ]}, where j represents the longitudinal area index of the carriage, including the front, middle and rear, k represents the transverse area index of the carriage, including the left and right, and x j,min , x j，max and k,min ,y k,max Represents the minimum and maximum boundaries of the partition in the x and y directions respectively;

[0221] The objective function of cargo distribution is defined as

[0222] Among them, M j，k Indicates area L j，k Current accumulated weight, V j，k Indicates area L j，k The current cumulative volume, ω1, ω2, ω3 represent the regional allocation weight coefficients, dist(p i , L j，k ) indicates the cargo g i Current location p i =(x i ,y i , z i ) to compartment L j，k distance;

[0223] g per piece of goods i Assign to target area

[0224] By priority di and regional allocation Sorting, generating optimized queue Q:

[0225] In step S4, loading is performed:

[0226] The mobile platform receives the cargo gi from the forklift and updates its coordinates to the current position of the platform Transfer to the retractable and foldable loading platform: in, represents the coordinates after transfer, T represents the coordinate transformation matrix from the mobile platform to the loading platform;

[0227] The retractable and foldable loading platform uses the forklift function to load the goods into the target area Adjust cargo position to optimize distribution: Among them, Δpi represents the offset between the current position of the goods and the center of the target area; adjust the position of the goods: in, is the final cargo location;

[0228] When the cargo is loaded to the rear and there is insufficient space in the carriage, the mobile platform moves backwards;

[0229] Specifically, in step S4, cargo analysis is performed, loading queue generation is optimized, and efficient and balanced loading is achieved in dynamic adjustment;

[0230] Step S5: platform reset.

[0231] After the loading task is completed, the auxiliary control center automatically returns to the initial position according to the path planning result of step S2.

[0232] Example 3, reference Figure 3-Figure 6 This embodiment provides a practical workflow of a safe intelligent assistance system and method for shipping machinery:

[0233] The mobile platform in the intelligent safety assistance system for loading machinery consists of: a retractable loading platform, a retractable and foldable loading platform, and a load-bearing walking mechanism:

[0234] Retractable loading platform: within a certain range, its height can be adjusted electrically, and can also be adjusted manually when there is no electric drive; the platform is equipped with electric-driven rollers to ensure that when a standard freight pallet is placed on it, it can move in both the front and rear directions, and its moving speed can be adjusted in multiple levels; when there is no electric drive, the freight pallet carried on it can also be moved manually, either by pushing it with hands or by cranking it, etc.

[0235] Retractable and foldable loading platform: within a certain range, its height can be adjusted electrically, and its relative position to the retractable loading platform can be adjusted electrically in the vertical direction; when there is no electric drive, the freight pallet carried on it can also be moved manually, either by pushing it with hands or by cranking it; this loading platform is equipped with a lifting device similar to a forklift, which can lift the goods on the platform out of the platform at the end of the platform and place them on other load-bearing surfaces;

[0236] Load-bearing walking mechanism: used to carry the retractable loading platform and the retractable and foldable loading platform, with rollers with a power device underneath, which can move in four horizontal directions under electric drive; it can also be moved manually in the absence of electricity; each wheel is equipped with a locking device, and after the final loading position is confirmed, the wheel is locked by the locking device to fix the position of the entire load-bearing walking mechanism; each wheel has a height adjustment function, and the height of each wheel is adjusted independently by hydraulics to ensure that the entire load-bearing walking mechanism is in a horizontal state.

[0237] like Figure 3 As shown in the figure, the nine gray rectangles around are the nine columns of the warehouse, which divide the warehouse space into four independent parts. The area on the right is the driveway. The loading truck needs to park in the upper part of this area, which is the "parking area", while the lower "loading area" is left vacant for the working area required by this system.

[0238] The "forklift" in the upper left corner is used in the left area of ​​the forklift (not shown in the picture) to take the goods to be loaded, and then turn left 90 degrees and lift them to the "retractable loading platform" in the lower left corner of the picture; the "retractable loading platform" and "retractable and foldable loading and unloading platform" in the lower left corner are both equipped with rolling pulleys, which can be driven to roll by electric means. In the absence of electric drive, the goods can also be moved by manual pushing; on the "retractable and foldable loading platform", there is also a lifting device similar to that on the forklift, that is, when the goods are moved to the front end of the platform, the lifting device can be used to lift the goods and put them directly into the car, thereby completing the loading of the goods;

[0239] Workflow:

[0240] Under normal circumstances, the "retractable loading platform" and the "retractable foldable loading platform" are in a vertical state, that is, after the "retractable foldable loading platform" flips 90 degrees to the "left", it is perpendicular to the "retractable loading platform", and the two components are in an "L" shape; the heights of the two platforms can be controlled independently, and the "retractable foldable loading platform" can also move forward and backward relative to the "retractable loading platform". However, no matter how it moves, the two devices are a whole, and they are on a device that can move "forward, backward, left and right"; the right lane is divided into two parts, namely the "parking area" above and the "loading area" below. Vehicles that need to be loaded, after entering from under the lane, should park in the "parking area" in the direction indicated by the arrow, and are required to align the rear of the vehicle to the front edge of the "loading area";

[0241] When a loading vehicle enters and stops in the "parking area", it enters the deployment state:

[0242] Reverse to the "right" and unfold the "retractable and foldable loading platform" so that it is in a "T" shape with the "retractable loading platform"; adjust the height of the two platforms so that they are relatively consistent with the height of the truck compartment and slightly higher than the truck compartment to facilitate the loading and unloading platform to extend into the truck compartment.

[0243] After accurately docking with the truck compartment, it enters the docking state:

[0244] The entire mobile platform including the "retractable loading platform" and the "retractable and foldable loading platform" moves toward the loading lane until the front edge of the "retractable and foldable loading platform" is aligned with the rear of the truck compartment; at this time, the position of the "retractable loading platform" is adjusted so that its upper edge touches the rear of the truck, slightly higher. The purpose of this movement is to ensure that the "retractable and foldable loading platform" is as deep as possible when it is moved into the truck compartment;

[0245] When the "retractable and foldable loading platform" is unfolded and inserted into the interior of the truck compartment, it enters the loading state:

[0246] After the "retractable and foldable loading platform" is in place, it continues to extend forward and into the interior of the vehicle, leaving enough "cargo placement" space in front of it; the forklift lifts the cargo to be shipped onto the "retractable loading platform", and then pushes the cargo horizontally to the "retractable and foldable platform" on the right through the cargo wheel on the platform, and then pushes the cargo upward to the front edge of the platform, and finally uses the forklift function on the loading platform;

[0247] Lift the goods into the carriage to complete the loading of the goods;

[0248] When the front part of the carriage is filled with cargo, the "retractable and foldable platform" retreats outward to make room for new cargo to wait for the next batch of cargo, and the "retractable and foldable loading platform" retreats while loading cargo until the entire carriage is full of cargo. During the loading process, the "retractable and foldable loading platform" can not only move forward and backward, but the entire system can also move left and right to ensure that the loading space of the entire carriage can be effectively utilized as soon as possible.

[0249] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A safe intelligent auxiliary system for loading and unloading machinery, comprising a mobile platform and an auxiliary control center, characterized in that: The mobile platform includes: a retractable loading platform, a retractable and foldable loading platform and a load-bearing walking mechanism; The auxiliary control center includes: environmental perception module, path planning module, dynamic load balancing module and multi-task collaborative scheduling module; The retractable loading platform can be adjusted in height electrically or manually, and the cargo can be moved electrically or manually; The retractable and foldable loading platform is electrically or manually adjustable in height, the relative position of the retractable loading platform is electrically or manually adjustable, and the carried goods are electrically or manually moved; Environmental perception module, which is used to perceive the work site environment in real time and generate dynamic 3D work maps to identify trucks, obstacles and loading and unloading area information; The path planning module is responsible for setting and executing the movement path of the mobile platform from the initial position to the truck docking point; Dynamic loading balance module, which controls the balance of the mobile platform during loading, monitors and adjusts the load distribution of the cargo; Multi-task collaborative scheduling module coordinates the allocation of work tasks between forklifts and platforms.

2. The intelligent safety assistance system for loading and unloading machinery according to claim 1 is characterized in that: The retractable and foldable loading platform integrates forklift function, which can lift the goods at the end for loading operation. The retractable and foldable loading platform integrates load pressure sensor network and hydraulic height adjustment unit; A network of load pressure sensors is distributed on the surface of the retractable and foldable loading platform to monitor the weight distribution, volume and location of the cargo in real time; A hydraulic height adjustment unit adjusts the height of each support point of the platform based on the output of the dynamic load balancing module; The load-bearing walking mechanism in the mobile platform supports the retractable loading platform and the retractable and foldable loading platform. It has powered rollers, can move in four horizontal directions, and has a locking device and hydraulic height adjustment function.

3. The intelligent safety assistance system for loading and unloading machinery according to claim 2 is characterized in that: The working states of the mobile platform include standby state, deployment state, docking state and loading state; Standby state: When the platform is not working, the retractable loading platform and the retractable and foldable loading platform maintain an L-shaped vertical state. After the truck stops in the parking area, it starts loading operations in the loading area and enters the unfolded state; Unfolded state: When the truck enters the loading area, the retractable loading platform and the retractable and foldable loading platform are unfolded to a T-shaped state, and the height of the retractable and foldable loading platform is adjusted to be consistent with the height of the truck compartment, and enters the docking state; Docking state: the mobile platform moves as a whole to the rear of the carriage, the retractable and foldable loading platform docks with the carriage, the retractable and foldable loading platform adjusts its position to go deep into the carriage, and enters the loading state; Loading status: The retractable and foldable loading platform goes further into the carriage and reserves enough space for the goods. The goods are delivered to the retractable loading platform by a forklift, rolled to the retractable and foldable loading platform, and finally lifted into the carriage by the forklift function of the retractable and foldable loading platform. The mobile platform keeps moving backwards before the carriage is full of goods to make room for new goods.

4. The intelligent safety assistance system for loading and unloading machinery according to claim 3 is characterized in that: The environmental perception module includes: LiDAR and cameras are used to generate a three-dimensional map of the site and identify obstacles in the warehouse environment; The truck positioning unit, combined with the optical marker or RFID tag at the rear of the truck, locates the specific position and docking point of the truck; The path planning module uses a reinforcement learning path planning algorithm to calculate the moving path of the mobile platform based on the 3D map of the work site. During the path execution process, the navigation path is adjusted according to the data of the environmental perception module to avoid obstacles; It is also responsible for docking the retractable and foldable loading platform to the rear of the truck.

5. The intelligent safety assistance system for loading and unloading machinery according to claim 4 is characterized in that: Dynamic load balancing module, which calculates the load distribution of the platform based on the cargo weight distribution monitored by the load pressure sensor network and outputs the height of the dynamic hydraulic height adjustment unit; The multi-task collaborative scheduling module determines the loading order of the goods based on the weight, volume and position of the goods obtained by the load pressure sensor network; and allocates the tasks of the forklift and the platform to ensure that there is no conflict between the two operations.

6. A method for intelligent assistance of loading machinery safety, characterized in that: A safe intelligent assistance system for loading machinery according to any one of claims 1 to 5 comprises the following steps: Step S1, environmental perception, Generate a real-time 3D map of the warehouse site based on LiDAR and cameras to identify obstacles in the loading and unloading area, the location of pillars, and the docking position of trucks; use optical markers at the rear of the carriage to locate the docking point of the truck, obtain a complete work site model and the location of the truck docking point; Step S2: dynamic path planning and initial positioning. Based on the 3D map in step S1, combined with the reinforcement learning path planning algorithm, a path is generated for the mobile platform from the current position to the truck docking point; In step S2, the mobile platform avoids obstacles along the way and adaptively adjusts the travel speed and direction. After reaching the target point, the relative position of the retractable and foldable loading platform and the carriage is adjusted so that the retractable and foldable loading platform docks with the rear of the carriage and enters the loading state. Step S3, dynamic load balancing optimization, Before loading, the dynamic load balancing module uses a balance optimization algorithm to calculate the cargo placement based on the cargo weight distribution monitored in real time by the load pressure sensor network, and then the height is adjusted by the hydraulic height adjustment unit; Step S4, multi-task collaborative loading, The multi-task collaborative scheduling module generates an optimized loading queue based on the cargo list and the spatial layout of the trucks: Prioritize loading heavy cargo to the bottom of the carriage or near the front of the carriage, and place larger cargo on both sides of the carriage; The mobile platform receives the cargo from the forklift and transfers the cargo to the retractable and foldable loading platform. The forklift function is used to load the cargo into the carriage one by one. When the cargo approaches the rear of the carriage, the mobile platform automatically moves back to release new loading space. Step S5: platform reset. After the loading task is completed, the auxiliary control center automatically returns to the initial position according to the path planning result of step S2.

7. A method for intelligent safety assistance of loading machinery according to claim 6, characterized in that: In step S1, the method of generating a 3D map of the warehouse site is as follows: Assume that the environmental point cloud data acquired by the laser radar is P L , P L ∈R N×3 , where N is the number of points in the point cloud, each point contains three-dimensional coordinates (x, y, z), R N×3 represents a matrix with N rows and 3 columns, each element is a real number, R represents the set of real numbers, P L ={(x i ,y i ,z i )|i=1,2,…,N}, where x i ,y i ,z i Represents the three-dimensional space coordinates of point i in the point cloud; The point cloud data is segmented on the ground, and the least square method is used to fit the plane equation: z = ax + by + c, where a, b, c represent the plane parameters obtained by the least square fitting of the point cloud data, and z represents the height of each point in the point cloud; Calculate the distance from the point to the plane and remove points whose distance is less than the threshold ∈ g The ground points are obtained to obtain the non-ground point cloud P′ L : P′ L ={(x i ,y i ,z i )|d i >∈ g }, where d i Represents the vertical distance from point i to the fitting plane, ∈ g Indicates the ground distance threshold; In step S1, depth information is generated based on the camera image: The camera captures the scene image I C , and use structured light or binocular vision to calculate the depth map: Among them, D C (x, y) represents the depth value corresponding to the pixel point (x, y), f represents the focal length of the camera, B represents the baseline length of the binocular camera, d l ,d r Indicates the disparity value of the corresponding pixel points of the left and right cameras; Convert the depth map to a point cloud P C : Among them, c x ,c y Indicates the camera focus coordinates, f x ,f y Represents the component of the camera focal length in the x and y directions; The point cloud is fused with the image data, and the camera point cloud P is transformed using the external parameter matrix T, including the rotation matrix R and the translation vector t. C Convert to the laser radar coordinate system to obtain the fused point cloud P F : P F =R·P C +t,P F =P′ L ∪P F , where R represents the rotation matrix from the camera coordinate system to the lidar coordinate system, and t represents the translation vector; For the fused point cloud P F Perform clustering and use the DBSCAN algorithm to separate obstacles, pillars, and other objects; In step S1, the truck docking point is located: Use the optical markers at the rear of the vehicle to locate the docking point, combined with the camera image I C Calculate the marker position: (x m ,y m ,z m )=R·(x p ,y p ,z p )+t, where (x p ,y p ,z p ) represents the three-dimensional coordinates of the optical marker in the camera coordinate system, (x m ,y m ,z m ) represents the coordinates of the optical marker in the laser radar coordinate system; According to the marking position (x m ,y m ,z m ) Calculate the position of the truck docking point relative to the mobile platform: Δ x =x m -x p ,Δ y =y m -y p ,Δ z =z m -z p , where Δ x , Δ y and Δ z Indicates the offset of the mobile platform relative to the truck docking point, (x p ,y p ,z p ) represents the current position of the mobile platform.

8. The method for intelligent safety assistance of loading machinery according to claim 7, characterized in that: In step S2, the path is generated based on the 3D map in the following manner: Define the state of the mobile platform as s t =(x t ,y t ,θ t ), indicating the position (x t ,y t ) and the orientation θ t , Define the action space as a t =(v t ,ω t ), where v t represents the linear velocity, ω t represents the angular velocity, Define the reward function R(s t ,a t )for: R(s t ,a t )=-α·d g -β·d o -γ·|ω t |, where d g Represents the Euclidean distance between the current position and the target point, d o It represents the distance between the current position and the nearest obstacle. α, β, and γ represent weight parameters, which balance the requirements of reaching the target point, avoiding obstacles, and smoothing the path. Select the deep reinforcement learning algorithm DQN and DDPG update strategy π(a t |s t ), optimize the expected cumulative reward: Where γ represents the discount factor, which controls the weight of future rewards, J(π) represents the expected cumulative reward of strategy π, E[·] is the expectation operator, and T represents the time step termination condition; In step S2, the path is adjusted in real time: The target path generated by path planning is expressed as That is, the target path generated by path planning is composed of N discrete points, and the two-dimensional plane coordinates of each point are (x i ,y i ), the mobile platform adjusts the speed and direction to track the path in real time, the linear velocity v t and angular velocity ω t for: v t =K v ·d t ,ω t =K ω ·(θ g -θ t ),in, Indicates the distance from the current position of the platform to the target point. Indicates the direction of the current target point, K v ,K ω represents the control gain coefficient, (x t ,y t ) represents the current position of the mobile platform, (x g ,y g ) represents the target point coordinates; If an obstacle is detected, the distance d from the obstacle to the platform is calculated o : If d o <∈ s , adjust the target direction: θ t =θ t +Δθ,Δθ=sgn(ω t )·K a ·(∈ s -d o ), where ∈ s Represents the safety distance threshold, K a represents the obstacle avoidance adjustment coefficient, (x o ,y o ) represents the position coordinates of the obstacle, P F is the fused point cloud data set; In step S2, the platform is docked with the rear of the carriage: When the platform reaches the target point, it enters the fine-tuning stage, adjusting the position of the retractable and foldable loading platform relative to the rear of the carriage and fine-tuning the offset: Δ x =x g -x t ,Δ y =y g -y t ,Δ z =z g -z t , where x g ,y g ,z g Indicates the coordinates of the docking point at the rear of the carriage, x t ,y t ,z t Indicates the current position of the retractable and foldable loading platform; Adjust the platform height h using the hydraulic unit t :h t =h g +K h ·(Δ z ), where h g Indicates the height of the docking point at the rear of the carriage, K h Indicates the height adjustment factor.

9. A method for intelligent safety assistance of loading machinery according to claim 8, characterized in that: In step S3, balance optimization is performed based on the real-time weight distribution: Assume that the sensor network of the loading platform consists of M×N distribution points, and the sensor is located at position (x i,j ,y i,j ) is recorded as f i,j , the total pressure matrix is ​​F: F = {f i,j |1≤i≤M,1≤j≤N}; Calculate the weight distribution of the sensor sampling points: Where g is the acceleration due to gravity, w is the i,j is the cargo weight at point (i, j); Calculate the total weight W and the center of gravity of the platform (x c ,y c ): In step S3, a balanced optimization algorithm is used to calculate the placement position: The goal is to make the center of gravity of the platform (x c ,y c ) approaches the target equilibrium point (x t ,y t ), objective function for: Optimize the placement of goods, assuming the weight of the new goods is w n , its candidate placement point is (x n ,y n ), optimize the model for: Enumerate candidate points (x n ,y n ) or gradient descent method to solve the optimal point; In step S3, the hydraulic height is adjusted in the following manner: The adjustment target is defined as: Keep the platform surface balanced so that the height of the hydraulic support point is h k,l In line with the optimized inclination requirements, the platform hydraulic support point is set as P k,l , whose coordinates are (x k,l ,y k,l ), the adjusted height h k,l for: h k,l =h0+K·(x k,l -x c )+K·(y k,l -y c ), where h0 represents the initial height of the platform, and K represents the adjustment coefficient, which is determined based on the calculation results of the dynamic loading balance module; According to the height difference Δh of the hydraulic support point k,l Adjust the final inclination angle of the platform: Among them, θ x ,θ y Indicates the inclination angle of the platform in the x and y directions.

10. A method for intelligent safety assistance of loading machinery according to claim 9, characterized in that: In step S4, the loading queue is adjusted in the following manner: Assume that the goods collection is g per piece of goods i The properties include: weight m i , volume v i , the initial position coordinates of the goods p i =(x i ,y i ,z i ) and loading priority d i ; Define loading priority according to carriage layout i : Where M represents the total weight of the cargo collection, V represents the total volume of the cargo collection, C represents the reference coordinates of the truck loading point, and dist(p i ,C) represents the distance between the initial position coordinates and the reference coordinates, α, β, γ represent the priority weight coefficients; In step S4, an optimized loading queue is generated: Divide the carriage into the front, middle, rear, and left and right sides, using the partition index L j,k express: L j,k ={(x,y,z)|x∈[x j,min ,x j,max ],y∈[y k,min ,y k,max ]}, where j represents the longitudinal area index of the carriage, including the front, middle and rear, k represents the transverse area index of the carriage, including the left and right, and x j,min ,x j,max and k,min ,y k,max Represents the minimum and maximum boundaries of the partition in the x and y directions respectively; The objective function of cargo distribution is defined as Among them, M j,k Indicates area L j,k Current accumulated weight, V j,k Indicates area L j,k The current cumulative volume, ω1, ω2, ω3 represent the area allocation weight coefficients, dist(p i ,L j,k ) indicates the cargo g i Current location p i =(x i ,y i ,z i ) to compartment L j,k distance; g per piece of goods i Assign to target area By priority d i and regional allocation Sorting, generating optimized queue Q: In step S4, loading is performed: Mobile platform receives cargo from forklift i , update its coordinates to the current position of the platform Transfer to the retractable and foldable loading platform: in, represents the coordinates after transfer, T represents the coordinate transformation matrix from the mobile platform to the loading platform; The retractable and foldable loading platform uses forklift function to load the goods into the target area Adjust cargo position to optimize distribution: Among them, Δp i Indicates the offset between the current position of the cargo and the center of the target area; adjust the cargo position: in, is the final cargo location; When the cargo is loaded to the rear and there is insufficient space in the carriage, the mobile platform moves backward.

Images