Tray-based trunk stacking planning method

Abstract

The invention relates to a trunk stacking planning method based on a tray, which comprises the following parts: an alignment mode: placing positions of a trunk on a gripper are left alignment and right alignment; the trunk placing method of the robot comprises the following three methods: the gripper directly moves to a target position, the gripper retreats, and the driving part on the gripper pushes the trunk forwards; when the gripper moves to the rear part of the target position, a driving part on the gripper forwards pushes the luggage case to the target position; when the luggage case falls down, the gripper moves to the position above the target position, the gripper retreats, the driving part on the gripper pushes the luggage case forwards, and the luggage case falls down to the target position. According to the actual scene, the hand grasp has three conditions, namely no offset, offset in the X direction or offset in the Z direction. And path selection: directly reaching the target point from the home point, or firstly reaching the intermediate point from the home point and then reaching the target point. According to the actual scene, the optimal combination scheme is selected for the trunk stacking position and path each time, so that the stacking efficiency, the space utilization rate and the like are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent palletizing of luggage boxes at airports, and in particular to a pallet-based luggage box palletizing planning method. Background Art

[0002] As the annual passenger throughput of airports grows rapidly, airports urgently need to improve overall operating efficiency through intelligent construction to ensure passengers' travel experience. Traditional airports still use manual operations to load airline baggage, which has low loading efficiency and high costs. At present, there are also a small number of airports trying to use intelligent luggage grabbing systems, including multi-axis robots, grippers, measurement sensors, label readers, palletizing planning algorithms, etc., to complete the grabbing of luggage from the operating assembly line and palletizing it on the transport vehicle through robots, which can reduce the luggage retention time in the sorting area, reduce manual labor intensity, and improve efficiency. However, due to the large differences in the shapes of luggage boxes at airports, there are differences in length, width and height. How to effectively plan the path for robots to transfer luggage and improve the space utilization of transport vehicles during palletizing is a problem that needs to be further solved. Summary of the invention

[0003] The purpose of the present invention is to provide a luggage box palletizing planning method based on a pallet, so as to improve the space utilization rate of a luggage box transport vehicle through a more reasonable palletizing planning method.

[0004] To achieve the above object, the present invention adopts the following technical solution:

[0005] A pallet-based luggage palletizing planning method includes the following parts: (1) Alignment: There are two positions for placing luggage on the gripper, namely left alignment and right alignment. The gripper is the robot end effector used to transfer the luggage;

[0006] The robot places the suitcase in three ways: (2-1) the gripper moves directly to the target position, the gripper moves backward, and the driving unit on the gripper pushes the suitcase forward; (2-2) the gripper moves to the back of the target position, and the driving unit on the gripper pushes the suitcase forward to the target position; (2-3) the gripper moves above the target position, the gripper moves backward, and the driving unit on the gripper pushes the suitcase forward, and the suitcase falls to the target position; each time the suitcase is placed, one of the above three methods is selected according to the actual scenario;

[0007] Offset: According to the actual scenario, there are three situations for hand grasping, namely no offset, offset in the X direction or offset in the Z direction;

[0008] Path selection: directly reach the destination point from the home point, or go from the home point to the intermediate point and then to the destination point.

[0009] Furthermore, the stacking of each suitcase is carried out through the planning of (1) to (4) above, and one of several specific planning schemes is selected.

[0010] Furthermore, a large number of planning schemes are counted, and based on the deep learning algorithm, the schemes with high success rates are preferentially selected from the feasible schemes.

[0011] Furthermore, the deep learning algorithm formula can be expressed as the output of neurons in the neural network and the gradient calculation in back propagation, and the formulas are:

[0012] (3) Output of neurons in a neural network

[0013] Assume that in the lth layer of the neural network, the output of neuron j is It can be expressed as:

[0014]

[0015] in:

[0016] g is the activation function;

[0017] is the weight between the jth neuron in layer l and the ith neuron in layer l-1;

[0018] is the bias term of the jth neuron in the lth layer;

[0019] n (l-1) is the number of neurons in the l-1th layer;

[0020] (4) Gradient calculation in back propagation

[0021] The back-propagation algorithm updates the network parameters by calculating the gradient of the loss function with respect to each weight and bias. For example, for weight Its gradient can be calculated using the chain rule:

[0022]

[0023] in:

[0024] L is the loss function;

[0025] is the error term of the jth neuron in the lth layer, defined as in is the weighted input of the jth neuron in the lth layer.

[0026] Furthermore, the hand grip adopts a tray, the tray has a supporting part for supporting the suitcase and limiting parts on the left and right sides, and a push plate or a conveyor belt is provided on the tray, and the push plate or the conveyor belt forms the driving part for pushing the suitcase in a straight line.

[0027] Furthermore, left alignment and right alignment refer to the side of the suitcase being aligned with the limiting parts on the left and right sides respectively.

[0028] Furthermore, based on the image acquisition of the luggage on the conveyor line by the visual camera, the measurement software calculates the luggage size information, which is then transmitted to the simulation software to generate the luggage model. The palletizing algorithm can calculate the luggage placement position, and then the simulation software plans the placement path, and the robot executes according to the path. The calculated luggage size information is the length, width, and height information of the luggage, i.e., l, w, h. Assumption: L i represents the size of the i-th suitcase, where L i =(l i ,w i ,h i ), respectively representing length, width, and height. i represents the placement of the i-th suitcase, where P i =(x i ,y i ), represents the coordinate position of the lower left corner of the suitcase on the plane. Assuming that the order of placement is from left to right, row by row, the calculation formula for the placement position can be simplified to: In actual applications, the placement position is not necessarily from left to right, and the simulation software needs to plan the placement path and calculate the optimal placement position of the currently grabbed suitcase.

[0029] Furthermore, by setting up a measuring camera in the palletizing area, an image of the palletizing area is acquired in real time to obtain the current pallet type information, and the placement position of the next suitcase is based on the pallet type information and the size of the suitcase.

[0030] Beneficial effects of the present invention:

[0031] The present invention provides a pallet-based luggage box palletizing planning method, which consists of the placement position of the luggage box on the pallet, the method of placing the luggage box, whether to offset, and path selection. According to the actual scenario (such as the size of the luggage box and the real-time space situation of the palletizing area), the optimal solution is selected to improve the palletizing efficiency, space utilization, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a schematic diagram of the core structure of the luggage box palletizing planning method based on the pallet of the present invention;

[0033] Figure 2It is a schematic diagram of a stacking state in the luggage box stacking planning method based on a pallet of the present invention;

[0034] Figure 3 It is another schematic diagram of stacking states in the luggage box stacking planning method based on a pallet of the present invention;

[0035] Figure 4 It is another schematic diagram of stacking states in the pallet-based luggage box stacking planning method of the present invention.

[0036] The names corresponding to the marks in the figure are:

[0037] 1. Palletizing area; 2. Pallet; 3. Luggage box A; 4. Luggage box B; 5. Luggage box C. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0039] Example:

[0040] The present invention provides a pallet-based luggage box stacking planning method, which mainly involves designing the placement position and transfer path of the luggage boxes. How a robot grabs the luggage boxes from a conveyor line is not an innovative content of the present invention.

[0041] Based on the image acquisition of the luggage on the conveyor line by the visual camera, the measurement software calculates the size of the luggage, which is then passed to the simulation software to generate a luggage model. The palletizing algorithm can calculate the placement of the luggage, and then the simulation software plans the placement path, and the robot executes according to the path. The palletizing position is determined by the size of the luggage to be palletized and the remaining space in the palletizing area. The palletizing area generally refers to the luggage transport vehicle, but is not limited to this.

[0042] The luggage palletizing planning method includes the following parts: Figure 1 As shown:

[0043] (1) Alignment: There are two positions for placing the suitcase on the gripper, namely left alignment and right alignment. The gripper is the robot's end effector and is used to transfer the suitcase to the target stacking position.

[0044] The hand grip uses a tray, which has a support part for supporting the suitcase and limit parts on the left and right sides to prevent the suitcase from falling off during the transfer process. Left alignment and right alignment refer to the side of the suitcase being aligned with the limit parts on the left and right sides respectively.

[0045] For ease of understanding, the following examples are given: Figure 2 As shown in the figure, when the target position of the suitcase palletizing is the leftmost side of the palletizing space, the position of the suitcase on the pallet is selected to be left-aligned. Figure 1 are the two states of suitcase A before and after it is placed. Similarly, Figure 3 As shown, if suitcase B needs to be placed next to suitcase A on the right side, suitcase B should also be left-aligned. On the contrary, if the suitcase to be placed is placed on the right side, the position of the suitcase on the tray should be right-aligned. As for left alignment and right alignment, it is determined according to the on-site conditions.

[0046] (2) Methods for placing a suitcase by the robot: There are three types: (2-1) the gripper moves directly to the target position, the gripper moves backward, and the driving part on the gripper pushes the suitcase forward.

[0047] A conveyor belt is provided on the pallet, and the transmission belt forms the driving part. When the transmission belt moves, the luggage box can be pushed straightly by friction, leaving the pallet, and the luggage box is placed at the target position of the stacking area. Figure 1 and Figure 2 This placement method can be used for all the palletizing positions shown.

[0048] The method of driving the luggage on the tray is not the key content of the present invention, and the implementation method in the prior art can be adopted. The transmission belt can be wound around the tray, and the axis of the rollers at both ends of the transmission belt is perpendicular to the pushing direction of the luggage, and the rollers on the rear side are driven by a servo motor. When necessary, the transmission belt can drive the luggage by the friction between the upper side of the transmission belt and the luggage. In other embodiments, a movable push plate can also be set on the tray as a driving part. The invention patent with publication number CN116573436A relates to an automatic sorting and loading device for luggage, which has an end effector and also uses a push plate to push the luggage off the tray.

[0049] (2-2) The gripper moves to the rear of the target position, and the driving unit on the gripper pushes the suitcase forward to the target position.

[0050] like Figure 4 As shown, the waiting position of suitcase C is the middle position, and other suitcases have been placed on the left and right sides. At this time, the space does not allow the tray to move to the target position (the tray is wider than the middle position), so it is chosen to move to the rear position and use the driving unit to push the suitcase forward.

[0051] (2-3) The gripper moves to the top of the target position, the gripper moves backward, and the driving unit on the gripper pushes the suitcase forward, and the suitcase falls to the target position.

[0052] against Figure 3 In the scenario shown (the middle position is reserved), if the suitcase is shorter, this method can be used, which allows the suitcase to be placed in place not only in the width direction but also in the rear, making full use of the stacking space.

[0053] Each time when placing luggage, one of the above three methods is selected according to the actual scenario. Through the measurement camera set up in the palletizing area, the image of the palletizing area is obtained in real time, and the current pallet type information is obtained in real time through software modeling. The placement of the next suitcase is based on the pallet type information and the size of the suitcase to be placed.

[0054] (3) Offset: Depending on the actual scenario, there are three situations for the hand grip, namely no offset, offset in the X direction or offset in the Z direction. The offset here refers to fine-tuning in the area where the suitcases are palletized to further improve the success rate of palletizing and prevent collisions, such as interference with already palletized suitcases. The X direction refers to the direction toward the robot (front and back direction), and the X direction refers to the vertical direction.

[0055] (4) Path selection: Go directly from the home point to the destination, or go from the home point to the intermediate point first and then to the destination. This is determined by the target location of the luggage stacking, which means that a reasonable path needs to be planned based on the actual scenario.

[0056] The luggage size information is defined as the length, width, and height of the luggage, i.e., l, w, h. Assumption: L i represents the size of the i-th suitcase, where L i =(l i ,w i ,h i ), respectively representing length, width, and height. i represents the placement of the i-th suitcase, where P i =(x i ,y i ), represents the coordinate position of the lower left corner of the suitcase on the plane. Assuming that the order of placement is from left to right, row by row, the calculation formula for the placement position can be simplified to: In actual applications, the placement position is not necessarily from left to right. The placement path needs to be planned through simulation software to calculate the optimal placement position of the currently grabbed suitcase.

[0057] The stacking of each suitcase is carried out according to the planning (1) to (4) above, and one of several specific planning schemes is selected.

[0058] Furthermore, a large number of planning schemes are counted, and based on the deep learning algorithm, the scheme with a high success rate is selected from the feasible schemes. Each of the planning steps (1) to (4) above has at least two options, so multiple schemes can be arranged and combined. For example, scheme 1: left-aligned, placed, not offset, and directly arrived; scheme 2: left-aligned, placed, not offset, and arrived at the middle point.

[0059] The core idea of ​​the above-mentioned deep learning algorithm is to learn the representation of data through a multi-level neural network structure to achieve complex pattern recognition and prediction tasks. The core is a multi-level neural network structure, which can learn multi-level feature representations through training data. The typical structure of a deep learning model includes an input layer, multiple hidden layers, and an output layer. Each hidden layer consists of multiple neurons (or nodes), each neuron is connected to the neurons in the previous layer through weights, and an activation function is applied to introduce nonlinearity. The training process of the entire network usually uses a backpropagation algorithm, combined with a gradient descent optimizer to adjust the weights so that the network's prediction results are as close to the actual labels as possible. The deep learning algorithm formula can be expressed as the output of neurons in the neural network and the gradient calculation in backpropagation. The formulas are:

[0060] (5) Output of neurons in a neural network

[0061] Assume that in the lth layer (hidden layer or output layer) of the neural network, the output of neuron j is It can be expressed as:

[0062]

[0063] in:

[0064] g is the activation function;

[0065] is the weight between the jth neuron in layer l and the ith neuron in layer l-1;

[0066] is the bias term of the jth neuron in the lth layer;

[0067] n (l-1) is the number of neurons in the l-1th layer.

[0068] (6) Gradient calculation in back propagation

[0069] The back-propagation algorithm updates the network parameters by calculating the gradient of the loss function with respect to each weight and bias. Its gradient can be calculated using the chain rule:

[0070]

[0071] in:

[0072] L is the loss function;

[0073] is the error term of the jth neuron in the lth layer, usually defined as in is the weighted input of the jth neuron in the lth layer. Through deep learning algorithms, a large amount of planning scheme data is used to learn complex patterns and relationships, thereby predicting the success rate of each scheme. Selecting a scheme with a high success rate can be achieved by setting a threshold, which can effectively help optimize planning decisions and resource allocation, and improve the success probability and efficiency of the scheme.

[0074] The present invention is not limited to the above-mentioned optimal implementation mode. Anyone can derive other various forms of products under the inspiration of the present invention. However, no matter what changes are made in the shape or structure, all technical solutions that are the same or similar to those of the present application fall within the protection scope of the present invention.

Claims

1. A luggage box palletizing planning method based on a pallet, characterized by: It includes the following parts,(1) Alignment: There are two positions for placing the suitcase on the gripper, i.e., left alignment and right alignment. The gripper is the robot end effector used to transfer the suitcase; (2) Methods for placing a suitcase by the robot: There are three methods: (2-1) the gripper moves directly to the target position, the gripper moves backward, and the driving unit on the gripper pushes the suitcase forward; (2-2) the gripper moves to the back of the target position, and the driving unit on the gripper pushes the suitcase forward to the target position; (2-3) The gripper moves to the top of the target position, the gripper moves backward, and the driving unit on the gripper pushes the suitcase forward, so that the suitcase falls to the target position. Each time the suitcase is placed, one of the above three methods is selected according to the actual scenario; (3) Offset: According to the actual scenario, there are three situations for hand grasping, namely, no offset, offset in the X direction, or offset in the Z direction; (4) Path selection: Go directly from the home point to the destination point, or go from the home point to the intermediate point and then to the destination point.

2. The luggage case palletizing planning method according to claim 1, characterized in that: The stacking of each suitcase is carried out according to the planning (1) to (4) above, and one of several specific planning schemes is selected.

3. The luggage case palletizing planning method according to claim 2, characterized in that: Count a large number of planning schemes, and based on deep learning algorithms, prioritize the schemes with high success rates from feasible schemes.

4. The luggage case palletizing planning method according to claim 1, characterized in that: The hand grip adopts a tray, which has a supporting part for supporting the suitcase and limiting parts on the left and right sides. A push plate or a conveyor belt is arranged on the tray, and the push plate or the conveyor belt forms the driving part for pushing the suitcase in a straight line.

5. The luggage case palletizing planning method according to claim 4, characterized in that: Left alignment and right alignment refer to the side of the suitcase being aligned with the limit parts on the left and right sides respectively.

6. The luggage case palletizing planning method according to claim 1, characterized in that: The visual camera collects images of the suitcases on the conveyor line, and the measurement software calculates the size of the suitcases. This information is then passed to the simulation software to generate a suitcase model. The palletizing algorithm can calculate the placement of the suitcases, and the simulation software plans the placement path, which the robot executes along.

7. The luggage case palletizing planning method according to claim 6, characterized in that: By setting up a measuring camera in the palletizing area, the image of the palletizing area is acquired in real time to obtain the current pallet type information. The placement of the next suitcase is based on the pallet type information and the size of the suitcase.

8. The luggage case palletizing planning method according to claim 3, characterized in that: The deep learning algorithm formula can be expressed as the output of neurons in the neural network and the gradient calculation in back propagation, and the formulas are: (1) Output of neurons in a neural network Assume that in the lth layer of the neural network, the output of neuron j is It can be expressed as: in: g is the activation function; is the weight between the jth neuron in layer l and the ith neuron in layer l-1; is the bias term of the jth neuron in the lth layer; n (l-1) is the number of neurons in the l-1th layer; (2) Gradient calculation in back propagation The back-propagation algorithm updates the network parameters by calculating the gradient of the loss function with respect to each weight and bias. For example, for weight Its gradient can be calculated using the chain rule: in: L is the loss function; is the error term of the jth neuron in the lth layer, defined as in is the weighted input of the jth neuron in the lth layer.

Images