CN109711624B - Boxing method, equipment and computer readable storage medium - Google Patents
Boxing method, equipment and computer readable storage medium Download PDFInfo
- Publication number
- CN109711624B CN109711624B CN201811626116.0A CN201811626116A CN109711624B CN 109711624 B CN109711624 B CN 109711624B CN 201811626116 A CN201811626116 A CN 201811626116A CN 109711624 B CN109711624 B CN 109711624B
- Authority
- CN
- China
- Prior art keywords
- goods
- boxing
- information
- support
- layer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Abstract
The invention discloses a boxing method, equipment and a computer readable storage medium, wherein the boxing method comprises the following steps: acquiring the containing information of the container and the packing information of goods to be packed; obtaining a boxing scheme according to the containing information, the boxing information and Monte Carlo tree search; in the process of obtaining a packing scheme according to the Monte Carlo tree search, a reward function is obtained according to the space utilization rate. The invention has the effect of improving the boxing universality and the calculation efficiency.
Description
Technical Field
The invention relates to the field of logistics sorting, in particular to a boxing method, equipment and a computer readable storage medium.
Background
Currently, boxing is usually performed manually according to experience. In order to improve the reasonableness and the efficiency of boxing, an intelligent algorithm can be adopted to assist in boxing.
The packing problem is a classical academic problem and has wide commercial application value. In the field of logistics, there is often the problem of the need to transport a given series of goods into a given container, such as a car.
Most existing algorithms can be divided into two categories, one is an optimization algorithm such as a genetic algorithm which is treated as a nonlinear optimization to solve by using a heuristic search algorithm which artificially defines rules, and the other is a deep learning algorithm.
However, the fundamental drawback of the first category of heuristic-based search methods is that their results rely on artificially specified heuristic rules. The result tends to be good when the rule applies, whereas the rule makes it difficult to derive a usable solution. However, most of the binning scenarios have complex constraints in their own right, which makes it difficult to find an applicable set of rules. Secondly, the rules have to reformulate the adjustments whenever there is a major change in the scene, which affects the generality of the algorithm itself.
The second category of methods, nonlinear optimization-genetic algorithm and deep learning algorithm, have two main drawbacks: and (one) lacks support for complex space physical constraints. These constraints make traditional algorithms difficult to handle; and (II) the mechanism of the nonlinear optimization algorithm is not supported by enough theory. Often requiring the user to spend a significant amount of time adjusting the parameters to achieve acceptable results. Therefore, the automation rate is low, and the popularization and the application of the algorithm in an industrial packing scene are restricted.
Disclosure of Invention
The invention mainly aims to provide a boxing method, equipment and a computer-readable storage medium, aiming at improving the boxing universality and the calculation efficiency.
In order to achieve the above object, the present invention provides a boxing method for logistics sorting, the boxing method comprising:
acquiring the containing information of the container and the packing information of goods to be packed;
obtaining a boxing scheme according to the containing information, the boxing information and Monte Carlo tree search;
in the process of obtaining a packing scheme according to the Monte Carlo tree search, a reward function is obtained according to the space utilization rate.
Optionally, the containing information includes the shape of the container and the shape and position information of the contained object;
the packing information comprises the shape, weight and bearing information of the goods to be packed;
the packing scheme includes information on the final position and orientation of each box in the container.
Optionally, the boxing method further includes:
in the simulation process of Monte Carlo tree search, under the condition that the expansion of a first layer tree node accords with the bearing limit, the expansion of a second layer tree node related to the tree node is carried out; otherwise, the expansion of another tree node of the first layer is executed.
Optionally, the boxing method further includes:
in the process of obtaining a packing scheme according to the Monte Carlo tree search, the stacking relation of the cargos is recorded through the AABB tree, and whether each cargo meets the bearing limit or not is judged.
Optionally, the recording, by the AABB tree, the stacking relationship of the cargos, and determining whether each cargo meets the load-bearing limit, includes:
when newly adding goods, obtaining a support object which is directly contacted with the bottom of the newly added goods;
and updating the bearing information of each support according to the weight of the newly added goods and the contact relation of the supports.
Optionally, the updating the weight bearing information of each support according to the weight of the newly added cargo and the contact relationship between the supports includes:
connecting the newly added goods to the upper layer of the supports to obtain the contact area of the newly added goods and each support and the total contact area of the newly added goods and all the supports;
when the weight or the bearing information of the goods on the upper layer changes, the support on the layer updates the bearing information according to the change; and the number of the first and second electrodes,
the support of the layer distributes the change according to the ratio of the contact area of the support and the goods of the previous layer to the total contact area of the goods of the previous layer; wherein the content of the first and second substances,
the support at the lowest layer is a support surface of the container.
Optionally, the boxing method further includes:
in the simulation process of Monte Carlo tree search, under the condition that the expansion of a tree node of a first layer accords with the space limitation, the expansion of a tree node of a second layer related to the tree node is carried out; otherwise, executing the expansion of another tree node of the first layer;
in the simulation process, the goods are placed in the discretization space, and whether the newly added goods accord with the space limitation or not is judged.
The invention also provides a boxing method for logistics sorting, which comprises the following steps:
obtaining the containing information of the container and the packing information of goods waiting to be loaded;
when the volume of the container is larger than a preset value, dividing the existing container into a plurality of areas;
according to the containing information and the boxing information, and according to Monte Carlo tree searching, obtaining each boxing proposal of each space one by one;
when searching for a packing scheme of a space, the containing information of the container is updated according to the previously obtained packing schemes of other spaces.
The invention also provides boxing equipment, which comprises a storage and a processor; the storage stores a boxing program;
the processor is used for executing the boxing program to execute the steps of the boxing method.
The invention also provides a computer readable storage medium having stored thereon a boxing program which, when executed by a processor, implements the steps of the boxing method as described above.
When the boxing method provided by the invention works, the preset space utilization rate is used as a reward function, and then the containing information of the container and the boxing information of the goods to be boxed are input. And then executing simulated packing by using the Monte Carlo tree search algorithm, namely continuously iterating the Action-State to obtain the direction of the simulated packing. And finally, calculating a reward function of the simulated boxing direction according to a reward function formula set according to the space utilization rate, selecting the optimal boxing direction according to a plurality of reward functions in different boxing directions, and then performing next iteration to obtain the next optimal boxing direction until all boxes are completely simulated, thereby obtaining the optimal boxing scheme. Therefore, the invention can solve the problem in a universal and automatic way through the Monte Carlo tree search algorithm, and solve the better packing solution between the goods to be packed and the container, thereby guiding the manual or robot to pack.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the structures shown in the drawings without creative efforts.
FIG. 1 is a flow chart of a first embodiment of the boxing method of the present invention;
FIG. 2 is a schematic diagram of a calculation process of the boxing method shown in FIG. 1;
FIG. 3 is a flow chart of a second embodiment of the boxing method of the present invention;
FIG. 4 is a flow chart of a third embodiment of the boxing method of the present invention;
FIG. 5 is a flow chart of a fourth embodiment of the boxing method of the present invention;
FIG. 6 is a flow chart of a fifth embodiment of the boxing method of the present invention;
FIG. 7 is a flow chart of a sixth embodiment of the boxing method of the present invention;
fig. 8 is a schematic diagram of a calculation process of the boxing method shown in fig. 7.
The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
In the specific embodiment, the monte carlo tree search method is used for performing packing simulation calculation, so that a series of goods with different shapes and physical characteristics are placed in a designated container, and certain space and physical limitation requirements are met.
For example: space requirements may include loading rates and other space constraints such as 10cm from the container, or the presence of already existing obstacles in the container like built-in corner fittings like containers. Physical requirements include box load-bearing regulations.
The load control may include that the actual load of a load in the structure must not exceed its nominal maximum load. Or to control the actual load bearing of the cargo to be no more than 90% of its nominal maximum load bearing, etc.
The process of computing a bin by the monte carlo tree search method includes:
first, the calculation process of the monte carlo tree search method is set.
The inputs to the algorithm may then be:
a. including the length, width and height of the target container,
b. shape and position information of an existing object in the target container can be included;
c. weight and weight bearing information of the existing object can also be included;
d. the information of the length, the width and the height of goods to be boxed is included;
e. the weight information and the bearing information of the goods to be boxed can also be included;
finally, the output of the algorithm may be:
f. including information on the location of each item within the container;
g. orientation information for each cargo may also be included.
After the Monte Carlo tree search calculation method is set, only the information of the container and the information of the container are needed to be input, and then the best packing scheme can be calculated. It should be noted that the calculated optimal solution is related to the reward function preset in the monte carlo tree search algorithm.
For example, if the space utilization of the container is set as the reward function, the packing plan obtained by calculation is the plan with the highest space utilization when the preset conditions (space or load/space and load) are satisfied. Of course, the load bearing utilization of the container can also be set as a reward function. The solution obtained at this time is the solution with the highest bearing utilization rate when the preset conditions (space or bearing/space and bearing) are met. Of course, it is also possible to provide a plurality of reward functions, for example, by simultaneously providing the space utilization and the load-bearing utilization of the container as reward functions, and these two elements each take five weight values.
The calculation process for the monte carlo tree search may include: simulating Action and State;
in the monte carlo tree search, each State corresponds to a structure existing in the current space, including goods already placed, and goods not yet placed.
The simulated Action in the monte carlo tree search may include information of the goods to be placed, where the goods information may include the material of the goods, the density of the goods, or the shape of the goods. Or in one embodiment the cargo information may also include the orientation of the cargo, the location where it is to be placed.
In the process of searching the Monte Carlo tree, an Action is executed in the original State, so that the box is put into a simulation mode under the condition that the space and bearing limits are met, and one box is put into the simulation mode every time the box is put into the simulation mode, namely, one layer is unfolded. Firstly, simulating to a preset expansion level, for example, simulating to put 10 boxes; or when the simulation is carried out until the last layer is unfolded, the last box to be placed is unfolded and placed. Then, the method returns to calculate which box the original State should be selected to be put in, i.e. which State1 substate in the first layer should be selected to be executed. The selection process is as follows: the calculation of the assigned reward function is performed for each State1 sub-State, and then the optimal one State1 sub-State is evaluated with the calculation result of the reward function. Finally, execution is determined to the State1 sub-State, making the State1 sub-State the new State State. Then the Action is executed again in a circulating way, and another new State is obtained. And (4) carrying out simulation and iteration repeatedly until the last box is put in or any box is put in virtually, wherein the set physical requirements cannot be met.
To summarize, the iterative operation of the monte carlo tree algorithm is divided into four steps:
the first step is as follows: selecting and selecting an original state. An executable tree node is selected, and a plurality of sub-nodes of the second layer are arranged below the tree node.
The second step is that: tree node expansion, i.e., expansion of tree nodes from the tree node S1 selected in the first step, results in a new child node S11. The sub-node S11 is one of the second level nodes connecting the nodes S1.
The third step: simulation, starting from sub-node S11, a subsequent random box-in operation is performed heuristically to obtain a subsequent deeper node. For example, the nodes S111, S112, S113 of the third layer of the simulation expansion, and so on; for example, the expanded fourth level nodes S1111, S1112, S1113, etc. are simulated from the node S111, and the expanded fourth level nodes S1121, S1122, S1123, etc. are simulated from the node S112.
In executing these simulation operations, it is necessary that the result of each simulation step satisfies the set physical constraints. The heuristic simulation randomly extracts the rest boxes, puts the boxes into the container one by one, and puts the number of the boxes until the space of the container can not be put into any box any more, or can not meet the physical constraint condition any more, or the maximum simulation spreading layer number supported by the system resources, and the like.
The fourth step: the reward function for the line at node S11 is evaluated and calculated according to the reward function formula.
The four steps are an iterative process.
And calculating the reward functions of the sub-nodes S11 of the second layer obtaining the original state, the reward functions of S12, the reward functions of S13 and the like at the same time of the preset number of iterations. And executing an optimal sub-node according to the reward function to form a new state.
If the space utilization rate is used as a reward function for judgment, iteration, in the same container, of which the total volume of the loaded goods accounts for the largest volume of the container serves as an optimal solution.
In executing the above-mentioned Action, regarding the calculation of the space limitation, the space of the container can be expressed by the discrete space expression method, and the position where the box can be placed is selected in the discretized space. For example, the discrete spatial representation may include: extreme point method Extreme Points or grid method Grids, and the like. Through discretization expression of the space, the continuous space can be changed into an array through sorting, the calculation amount can be reduced, and the calculation efficiency can be increased.
In executing the above-mentioned Action, the calculation of the bearing limit can record the up-down relative relationship of each cargo through the AABB-Tree structure. Then when each Action is taken, the newly added goods of the Action are added into the AABB-Tree structure, and the new load bearing is calculated.
In order to further reduce the amount of calculation, it is also possible to adjust the volume of the container. This problem is solved, for example, by the strategy of Divide-and-Conquer, namely:
divide-decompose the problem into sub-problems of smaller scale;
treating- -breaking through these smaller sub-problems one by one;
and merging, merging the solved subproblems, and finally obtaining the solution of the 'mother' problem.
In implementation, the container may be divided into a plurality of regions, then the optimal solution of each region is calculated, and finally all the optimal solutions are combined, so as to obtain the boxed optimal solution of the whole container.
Based on the idea of realizing the packing by the Monte Carlo tree searching method, the invention provides various embodiments of the method.
Example one
This embodiment provides a first embodiment of a method of boxing.
Referring to fig. 1, a boxing method for logistics sorting comprises:
s101, obtaining containing information of a container and packing information of goods to be packed;
step S102, a boxing scheme is obtained according to the containing information, the boxing information and Monte Carlo tree searching;
and step S103, obtaining a reward function according to the space utilization rate in the process of obtaining the packing scheme according to the Monte Carlo tree search.
In the present embodiment, the containing information of the container and the packing information of the goods to be packed are first obtained. Wherein, the containing information of the container comprises the length, the width and the height of the container and the information of the articles which are already in the container. Such as by corner information, or box information already placed inside, etc. The packing information of the goods to be packed may include ID, weight, volume, whether there is a requirement for placing the goods upward, whether there is a load limitation, and the like.
In this embodiment, after the containing information and the boxing information are obtained, a boxing scheme is obtained according to the containing information and the boxing information and according to a monte carlo tree search. Wherein the containment information and the binning information are input information. The calculation method of the Monte Carlo tree search is to continuously simulate the placement of boxes according to random numbers. And finally obtaining the final and optimal state, namely the boxing scheme, through the calculation process of Monte Carlo tree search.
In the present embodiment, in the process of obtaining a packing scheme according to the monte carlo tree search, a reward function is obtained according to the space utilization. The method comprises the steps of obtaining a reward function searched by a Monte Carlo tree according to space utilization rate, evaluating each iteration in a simulated boxing process, and finally taking a scheme formed by a series of optimal iterations as a boxing scheme. For example, the space utilization rate of different nodes of the same layer, such as 80% -98%, 70% -95%, 85% -92%, 70% -80% and the like, can be obtained by calculating the ratio between the total volume of the loaded boxes and the volume of the container; wherein, the highest space utilization rate is 98%, the nodes corresponding to 80% -98% space utilization rate can be used as the virtual boxing direction, so as to obtain a new State; starting from the new State, obtaining a node with the highest space utilization rate in the next layer, wherein the node is used as another new State; finally, in a series of optimal choices, a boxing scheme is obtained.
After the boxing scheme is obtained, a human or robot may be controlled to load boxes into containers one by one according to the boxing scheme.
According to the boxing method provided by the embodiment, when the boxing method works, the reward function is obtained through the preset space utilization rate, and then the containing information of the container and the boxing information of goods to be boxed are input. And then carrying out simulated packing by the Monte Carlo tree search algorithm, namely continuously carrying out Action-State iteration to obtain the simulated packing direction. And finally, calculating the reward function of the simulated packing direction by setting a reward function formula according to the space utilization rate. And selecting the optimal boxing direction according to a plurality of reward functions in different boxing directions, and then performing the next iteration to obtain the next optimal boxing direction until all boxes are completely packed in a simulation manner, thereby obtaining the optimal boxing scheme. Therefore, the embodiment can solve the packing optimal solution between the goods to be packed and the container through the Monte Carlo tree search algorithm in a universal and automatic way, so as to guide manual or robot to pack.
Further, in the present embodiment, the containing information includes a shape of the container and shape and position information of the contained object. Therefore, the boxing method provided by the embodiment can be used for manually putting part of important goods into the container and then calculating. Or perform relay calculations, etc.
Further, in this embodiment, the packing information includes information on the shape, weight, and bearing weight of the goods to be packed. Therefore, the boxing method provided by the embodiment can calculate the bearing limit of each Action according to the bearing information of each cargo.
Further, in this embodiment, the packing scheme includes information on the final position and orientation of each box in the container. Therefore, the boxing method provided by the embodiment can simulate the direction of the goods in different directions when each Action is performed, and increase the diversity of boxing, so that the result obtained by simulation is more real.
Further, in this embodiment, the boxing method further includes:
step S104, in the simulation process of Monte Carlo tree search, under the condition that the expansion of a tree node of a first layer accords with the bearing limit, the expansion of a tree node of a second layer related to the tree node is carried out; otherwise, the expansion of another tree node of the first layer is executed. The first layer and the second layer merely represent the relationship between the upper and lower layers, and do not limit the specific layers in the whole.
As shown in fig. 2, there are now 5 boxes that need to be placed in the container. Wherein:
in the State00000, the container is empty, and the goods to be loaded are required to be loaded into the container. At this time, the goods need to be loaded at a first position, which is a predetermined position, such as a corner designated on the inner side of the container. There are currently 5 items to be loaded. Each box can have six orientations. Therefore, when the first position is installed, one of the sub-nodes of 5 × 6 to 30 may be arbitrarily expanded starting from the State00000 root node. That is, boxes 1, 2, 3, 4, or 5 are placed into designated corners of the container, each of which can be placed in any of six positions.
Here, in order to make the explanation simpler, it is set that each cargo can be placed only in one posture. Thus, when installed in the first position, the expandable sub-nodes are reduced to 5.
Starting from the State00000 root node, the first iteration starts: randomly selecting any sub-node of the 5 sub-nodes to expand, simulating and expanding subsequent nodes, and judging the condition of bearing limit when one node is expanded. If the sub-nodes expanded by simulation meet the bearing limit, any sub-node of the sub-node can be expanded continuously, so that simulation is executed continuously. When the simulation is executed until no sub-node can meet the bearing limit or all boxes are put, calculating and returning to the reward function of the sub-node which is unfolded first in the first iteration. And starting from the State00000 root node, starting second iteration, optionally expanding one of the rest 4 sub-nodes, simulating, calculating and returning a reward function, and thus finishing the second iteration. Until reaching the iteration of the preset times, or 5 iterations are performed, namely the reward functions of 5 sub-nodes of the State00000 root node are calculated.
If the value of the reward function of the first placement box 4 is calculated to be the maximum at this time, the first placement box 4 is selected, and a new State, namely the State40000, is formed.
In the State40000 State, case 1, case 2, case 3, and case 5 can be loaded into the container at this time, and the positions that can be loaded include the second position, the third position, the fourth position, and the fifth position, which will be difficult to explain due to an excessive amount of change. For the sake of convenience of explanation, the position of the container is fixed at this time. That is, subsequent actions are placed in the second position, the third position, the fourth position and the fifth position in sequence.
And then, 4 times of iteration can be performed in the State of State40000, and if the iteration process of putting the tree nodes into the box 1 is performed, the situation that the expansion of the nodes of a certain layer of tree does not meet the bearing limit is simulated. The iteration with respect to box 1 is interrupted and the iteration of placing box 2 is performed again. If the iteration process of putting the tree into the box 2 is executed, the condition that the expansion of a certain layer of tree nodes does not meet the bearing limit is simulated. The iteration with respect to box 2 is interrupted and the iteration of placing box 3 is performed again. In the iteration process of putting the box 3 and the iteration process of putting the box 5, each simulation meets the bearing limit, and then the reward functions of putting the box 3 and putting the box 5 are obtained. In this case, the value of the reward function put in the box 3 is considered to be larger by comparison. Then the second position is selected for placement in box 3 and a new State is formed, State 43000.
By analogy, in the subsequent simulation and iteration processes, the third position is selected to be placed into the box 2, and a new State-State 43200 is formed. The fourth position is selected to place box 5, the fifth position to place box 1, and finally a new State-State 43251 is formed.
According to the packing method provided by the embodiment, the load-bearing limit is calculated through each simulation step in each iteration process, so that subsequent simulation can be continued only when the simulation which meets the load-bearing limit is performed, the calculation amount of Monte Carlo tree searching is reduced, the algorithm efficiency is improved, and the waiting time is shortened.
Of course, in other embodiments, it may be determined whether the simulation in the iteration meets the load-bearing limit after one iteration in the monte carlo tree search is completed. For example, in the above example, after the iteration for the first placing box 4 is completed, the overlapped structure in which the box 4, the box 3, the box 2, the box 5, and the box 1 are placed in order is constructed, and then the judgment of the load bearing limit is performed, and the judgment result is obtained as the load bearing limit is met. And after the other iteration is finished, constructing an overlapped structure which is sequentially placed into the box 4, the box 2, the box 3, the box 5 and the box 1, judging the bearing limit, and obtaining a judgment result that the judgment result is that the bearing limit is not met. The simulation that meets the load bearing limit participates in the earning of the reward function, while the simulation that does not meet the load bearing limit does not participate in the earning of the reward function.
Example two
Referring to fig. 3, the present embodiment provides a boxing method. The present embodiment is based on the above embodiments, and additionally adds steps. The method comprises the following specific steps:
the boxing method further comprises the following steps:
step S205, in the process of obtaining the packing scheme according to the Monte Carlo tree search, the stacking relation of the cargos is recorded through the AABB tree, and whether each cargo meets the bearing limit or not is judged.
Other steps of this embodiment are the same as those of the above embodiment, and reference may be made to the above embodiment for details, which are not described herein again.
In this embodiment, in the process of obtaining a packing scheme according to the monte carlo tree search, the stacking relationship of the cargos is recorded by the AABB tree, and it is determined whether each cargo meets the load bearing limit. Among them, the AABB (Axis Aligned Bounding Box) structure is a minimum rectangular parallelepiped enclosing an object parallel to coordinate axes. The AABB tree is a hierarchical binary tree constructed based on an AABB structure. And recording the stacking relation of the boxes through the AABB tree, and adding the newly added goods to the AABB tree to generate a new node when simulation is performed. Whether each cargo meets the load-bearing limit is then calculated through the stacking relationship between the cargos in the AABB tree.
In this embodiment, during the simulation, an AABB tree may be introduced and the bearing determination may be performed. So that simulations that do not comply with the load bearing limit will be interrupted in time, thereby reducing the amount of calculations and obtaining iteration results without the risk of not complying with the load bearing limit. Specifically, when each tree node is unfolded, the bearing judgment is performed on the state corresponding to the tree node; or, when the two-step tree node expansion is executed, the bearing judgment is carried out on the state corresponding to the tree node expanded in the second step; or, when the tree node expansion in the N steps is executed, the bearing judgment is performed on the state corresponding to the tree node expanded in the N step.
In this embodiment, whether each cargo in the simulated container meets the load-bearing limit is judged again by recording the stacking relationship of the cargo by adopting the AABB tree. Therefore, the description method of the AABB tree is utilized, so that the goods are more convenient to mathematically process, and the load and geometric splicing (the geometric splicing is a collision test, and the box is simulated through a collision model, for example, when a gap exists between the box A and the box B, a box C is required to be inserted, but when the length or the width of the box C is larger than the gap, the situation that the box C cannot be placed between the box A and the box B can be obtained through the collision test) is also more convenient to traverse the load calculation by utilizing the property of the binary tree. Therefore, the method has the effects of simple, stable and efficient calculation process.
EXAMPLE III
Referring to fig. 4, the present embodiment provides a boxing method. This example is based on the above examples, and further illustrates the steps therein. The method comprises the following specific steps:
the recording of the stacking relationship of the goods through the AABB tree and the judgment of whether each goods meets the bearing limit include:
step S301, when a new cargo is added, a support object directly contacted with the bottom of the new cargo is obtained;
and S302, updating the bearing information of each support according to the weight of the newly added goods and the contact relation of the supports.
Other steps of this embodiment are the same as those of the above embodiment, and reference may be made to the above embodiment for details, which are not described herein again.
In this embodiment, when the nodes are expanded in each simulation in the iterative process, that is, when the new cargo is added, the bottom of the new cargo is obtained to directly contact the support. Wherein, each time emulation all relates to newly-increased goods serial number, this newly-increased goods volume, weight and bearing information, the position information that the goods will be put promptly etc.. The support directly contacted with the bottom of the newly added cargo is obtained through analysis, and the weight of the newly added cargo directly influences the bearing information of the support.
In this embodiment, after the information of the supports is obtained, the weight bearing information of each support is updated according to the weight of the newly added cargo and the contact relationship of the supports. Wherein, the contact relationship can be whether the box in direct contact is the main support or the auxiliary support of the newly added box? The ratio of the area of support in direct contact with the box to the total area of support, etc. Therefore, when adding new goods, the corresponding weight sharing relationship can be obtained through the contact relationship. And then updating the weight of the newly added goods to each support according to the weight sharing relation. Since the stacking relationship is recorded through the AABB tree, when each cargo is recorded, the load bearing limit of the cargo and the current load bearing information are also recorded. So that it is only necessary to add a new share of the weight to each support after the weight sharing relationship of each box is obtained.
I.e. the weight bearing information of the directly contacting support is updated according to the weight of the newly added cargo. Only the updating of the load bearing information of the directly contacting support is described here. It will be appreciated that if there are other supports beneath the support that the support is in direct contact with, then the change in load bearing information for that support will also be transmitted to the other supports. Therefore, with continuous two-stage conduction, the weight of the newly added goods is updated to the tree root all the time, and therefore updating of all bearing information is completed.
According to the boxing method disclosed by the embodiment, the AABB tree executes bearing updating during simulation execution, so that whether the simulation meets the bearing limit can be judged in time, the simulation which does not meet the bearing limit can be interrupted in time to continue execution, the integral calculation amount is reduced, the system resources are saved, and the waiting time is shortened.
Further, the AABB tree is optimized by adopting a representation method of a non-connected directed graph. Specifically, each cargo is a node in the graph, and the edge from node a to node B represents a support B. By the non-communicated directed graph representation method, the algorithm can be used for updating the box body bearing quickly through breadth-first traversal.
Example four
Referring to fig. 5, the present embodiment provides a boxing method. This example is based on the above examples, and further illustrates the steps therein. The method comprises the following specific steps:
the updating of the bearing information of each support according to the weight of the newly added cargo and the contact relationship of the directly contacted cargo comprises:
step S401, connecting the newly added goods to the upper layer of the supports, and obtaining the contact area of the newly added goods and each support and the total contact area of the newly added goods and all the supports;
step S402, when the weight or the bearing information of the goods on the upper layer changes, the bearing information of the support on the layer is updated according to the change; and the number of the first and second electrodes,
the support of the layer distributes the change according to the ratio of the contact area of the support and the goods of the previous layer to the total contact area of the goods of the previous layer; wherein the content of the first and second substances,
the support at the lowest layer is a support surface of the container.
Other steps of this embodiment are the same as those of the above embodiment, and reference may be made to the above embodiment for details, which are not described herein again.
In this embodiment, when performing simulation, the newly added cargo is attached to the upper layer of the supports, and the contact area of the newly added cargo with each of the supports and the total contact area with all the supports are obtained. Wherein the newly added cargo is attached to the upper layer of the directly contacting supports in the AABB tree. When other goods are connected to the newly added goods, the load of the newly added goods is changed under the influence of the other goods, so that the newly added goods are changedAnd the bearing information of the support positioned at the lower layer of the newly added goods correspondingly changes. The contact area is the area supported on the ground for the newly added goods. For example, two supports are directly contacted with each other under the newly added goods, and the contact area of each support is 10cm2The total contact area is 20cm2(ii) a The newly added goods are used as the upper layer of the two supports below.
In the present embodiment, after obtaining the contact area and the total contact area, the following is performed: when the weight or the bearing information of the goods on the upper layer changes, the support on the layer updates the bearing information according to the change; and the support of the layer distributes the change according to the ratio of the contact area of the support of the layer to the contact area of the load of the upper layer to the total contact area of the load of the upper layer, wherein the support of the lowest layer is the support surface of the container. The condition that the weight of the goods on the upper layer changes is suitable for the situation that the newly added goods are added on the upper layer of the support. The condition that the bearing information of the goods on the upper layer changes is suitable for the condition that the bearing information of the goods on the upper layer of the support changes. That is, the weight of the newly added goods directly changes the bearing information of the next layer of goods on the tree and indirectly changes the bearing information of the goods on the lower layer of the tree; the process of the influence is carried out in a layer-by-layer transmission mode and is transmitted to the position of the tree root.
Specifically, the weight proportion is as follows: if a new load is added to the upper layer of a support or the load bearing information of the load on the upper layer is changed, the weight or the changed load bearing information is transmitted to the support. Firstly, obtaining the contact area of a support and the goods on the upper layer and the contact area of the goods on the upper layer and all the supported goods; the support then distributes the weight or weight bearing information variation of the load in a layer above the contact area to the total contact area.
For example:
as shown in fig. 8, the new cargo a is placed on the upper layer of the support H, and the weight of the new cargo a is 10 kg; wherein the contact area of the support H contacting the newly added goods A is 10 square centimeters, and the total contact area of the newly added goods A and all the supports T on the lower layer is 50 square centimeters; thus, the support H will share a weight of 10kg × 10/50-2 kg. The 2kg weight will be added to the load bearing information of the support. While the remaining 8kg will be distributed to the other supports.
The load bearing information of the upper layer of the goods on the support H is changed, so that the weight of the support H is increased by 2 kg; wherein, the direct contact area of the support H and the next layer of goods B is 40 square centimeters, and the total contact area of the support H and all the supports on the lower layer is 50 square centimeters; thus, the next layer of cargo B will be allocated a weight of 2kg × 40/50-1.6 kg. This 1.6kg weight will be added to the load bearing information of the next layer of cargo B. While the remaining 0.4kg will be spread out on the other supports of the lower layer.
In the boxing method provided by the embodiment, when new goods are added, the new goods are established on the AABB tree and are established on the upper layer of the support. Then, when weight and bearing information change, the weight is transmitted and distributed layer by layer according to the sequence from top to bottom until the weight is distributed to the position of the tree root. Finally, the supporting surface of the container is taken as the lowest layer of support, namely the position of the tree root. The entire weight will be recorded onto the support surface of the container to determine if the weight of the entire load exceeds the load bearing limit of the container. Therefore, the boxing method provided by the embodiment distributes the weight layer by adopting a mode of supporting area ratio, and has the effects of reasonable weight distribution and high distribution calculation speed.
EXAMPLE five
Referring to fig. 6, the present embodiment provides a boxing method. The present embodiment is based on the above embodiments, and additionally adds steps. The method comprises the following specific steps:
the boxing method further comprises the following steps:
step S501, in the simulation process of Monte Carlo tree search, under the condition that the expansion of a tree node of a first layer meets the space limitation, the expansion of a tree node of a second layer related to the tree node is carried out; otherwise, executing the expansion of another tree node of the first layer;
step S502, in the simulation process, the goods are placed in the discretization space, and whether the newly added goods accord with the space limitation or not is judged.
Other steps of this embodiment are the same as those of the above embodiment, and reference may be made to the above embodiment for details, which are not described herein again.
In this embodiment, in the simulation process of performing the monte carlo tree search, when the expansion of a tree node of the first layer meets the space limitation, the expansion generation of a tree node of the second layer related to the tree node is performed; otherwise, the expansion of another tree node of the first layer is executed. The limitation of the space can be judged by judging whether the length, width and height of the accumulated goods are higher than the length, width and height of the volume. In the process of simulating and unfolding a tree node each time, judging the length, the width and the height of the goods; if the space constraint is met, the next tree node can be expanded in a simulation mode. If the space constraint is not satisfied, the expanded tree node is discarded and other tree node expansions are performed.
In this embodiment, in the simulation process, when it is determined whether the space limitation is met, the goods are placed in the discretized space, and it is determined whether the newly added goods meet the space limitation. The space is expressed in a discretization mode, so that the occupation of resources can be reduced, calculation can be simplified, and the calculation efficiency is improved.
EXAMPLE six
The present embodiment provides a method of boxing.
Referring to fig. 7, a boxing method for logistics sorting includes:
step S601, obtaining the containing information of the container and the packing information of the goods waiting to be loaded;
step S602, when the volume of the container is larger than a preset value, dividing the existing container into a plurality of areas;
step S603, obtaining each packing scheme of each space one by one according to the containing information, the packing information and Monte Carlo tree search;
in step S604, when the packing plan of one space is searched, the container information is updated according to the previously obtained packing plan of the other space.
In the present embodiment, the containing information of the container and the packing information of the goods waiting to be loaded are first obtained. As described above, the storage information of the container includes the length, width, and height of the container, and the information of the article already present inside the container. Such as by corner information, or box information already placed inside, etc. The packing information of the goods to be packed may include ID, weight, volume, whether there is a requirement for placing the goods upward, whether there is a load limitation, and the like.
In this embodiment, after the containing information and the packing information are obtained, the existing container is divided into a plurality of regions when the volume of the container is greater than a preset value. The preset value may be preset, for example, 1m3 or 2m3 when the preset value is 0.2m 3. When the volume exceeds a preset value, the container can be divided into two areas, three areas or four areas, and the like.
In this embodiment, after the container is divided into a plurality of areas, the container solutions of the spaces are obtained one by one according to the containing information and the packing information and according to the monte carlo tree search. When searching according to the Monte Carlo tree, the reward function can be obtained according to the space utilization rate of the container and/or the bearing utilization rate of the container. And according to the judgment of the space limit and the bearing limit in the Monte Carlo tree searching process, the Monte Carlo tree searching process is in accordance with the space limit and/or the bearing limit. For a specific way of judging the space limitation and the bearing limitation, refer to the above embodiments, and are not described herein again. After searching through the monte carlo tree, the binning scheme for any one region can be output. As shown in fig. 8, the container is divided into two sections, and section 1 is first loaded and section 1 is filled.
In this embodiment, when a packing plan of one area is obtained and then a packing plan of another space is searched, the container accommodation information is updated based on the previously obtained packing plan of the other space. Wherein, the collection of the packing schemes of other areas obtained before is used as the starting State in the new space calculation. As shown in fig. 8, when the area 1 is full, the packing of the area 2 is calculated again, and when the area 1 is full, the packing is calculated as the start State of the area 2.
According to the packing method provided by the embodiment, the large container is divided to form the plurality of small areas, and then each small area is packed one by one, so that the random operation amount during Monte Carlo tree searching is avoided, and the overlong calculation time of each step caused by overlarge selectable space is avoided.
EXAMPLE seven
The present embodiment provides a boxing apparatus.
The boxing apparatus comprises a storage and a processor; the storage stores a boxing program;
the processor is configured to execute the boxing program to perform the steps of the boxing method according to any one of the above embodiments.
Since the present embodiment has all the technical features of the above-mentioned boxing method, the present embodiment also has the beneficial effects of the above-mentioned boxing method. Please refer to the above embodiments, which are not described herein.
Example eight
The present embodiment provides a computer-readable storage medium.
The computer readable storage medium has stored thereon a binning program for execution by a processor of the steps of a binning method as described in any of the above.
Since the present embodiment has all the technical features of the above-mentioned boxing method, the present embodiment also has the beneficial effects of the above-mentioned boxing method. Please refer to the above embodiments, which are not described herein.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal (such as a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present invention.
While the present invention has been described with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, which are illustrative and not restrictive, and it will be apparent to those skilled in the art that various changes and modifications can be made therein without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (7)
1. A boxing method for logistics sorting, characterized in that the boxing method comprises:
acquiring the containing information of the container and the packing information of goods to be packed;
obtaining a boxing scheme according to the containing information, the boxing information and Monte Carlo tree search;
in the process of obtaining a boxing scheme according to Monte Carlo tree searching, obtaining a reward function according to the space utilization rate;
the method for obtaining the boxing scheme according to the Monte Carlo tree search specifically comprises the following steps:
continuously simulating and placing boxes according to random numbers, obtaining a reward function searched by a Monte Carlo tree according to the space utilization rate, evaluating each iteration in the simulated boxing process, and finally taking a scheme formed by a series of optimal iterations as a boxing scheme;
in the process of obtaining a packing scheme according to the Monte Carlo tree search, recording the stacking relation of the cargos through an AABB tree, and judging whether each cargo meets the bearing limit;
the recording of the stacking relationship of the goods through the AABB tree and the judgment of whether each goods meets the bearing limit include:
when newly adding goods, obtaining a support object which is directly contacted with the bottom of the newly added goods;
updating the bearing information of each support according to the contact relationship between the weight of the newly added goods and the supports;
the updating of the bearing information of each support according to the weight of the newly added cargo and the contact relationship of the supports comprises:
connecting the newly added goods to the upper layer of the supports to obtain the contact area of the newly added goods and each support and the total contact area of the newly added goods and all the supports;
when the weight or the bearing information of the goods on the upper layer changes, the support on the layer updates the bearing information according to the change; and the number of the first and second electrodes,
the support of the layer distributes the change according to the ratio of the contact area of the support and the goods of the previous layer to the total contact area of the goods of the previous layer; wherein the content of the first and second substances,
the support at the lowest layer is a support surface of the container.
2. A boxing method in accordance with claim 1,
the containing information includes a shape of the container and shape and position information of the contained object;
the packing information comprises the shape, weight and bearing information of the goods to be packed;
the packing scheme includes information on the final position and orientation of each box in the container.
3. A boxing method in accordance with claim 2, wherein the boxing method further comprises:
in the simulation process of Monte Carlo tree search, under the condition that the expansion of a first layer tree node accords with the bearing limit, the expansion of a second layer tree node related to the tree node is carried out; otherwise, the expansion of another tree node of the first layer is executed.
4. A boxing method in accordance with claim 2, wherein the boxing method further comprises:
in the simulation process of Monte Carlo tree search, under the condition that the expansion of a tree node of a first layer accords with the space limitation, the expansion of a tree node of a second layer related to the tree node is carried out; otherwise, executing the expansion of another tree node of the first layer;
in the simulation process, the goods are placed in the discretization space, and whether the newly added goods accord with the space limitation or not is judged.
5. A boxing method for logistics sorting, characterized in that the boxing method comprises:
obtaining the containing information of the container and the packing information of goods waiting to be loaded;
when the volume of the container is larger than a preset value, dividing the existing container into a plurality of areas;
according to the containing information and the boxing information, and according to Monte Carlo tree searching, obtaining each boxing proposal of each space one by one;
updating the containing information of the container according to the previously obtained packing schemes of other spaces when searching for the packing scheme of a space;
the container loading schemes in each space specifically comprise the following steps of obtaining the container loading information and the container loading information one by one according to Monte Carlo tree searching, wherein the container loading schemes in each space specifically comprise:
continuously simulating and placing boxes according to random numbers, obtaining a reward function searched by a Monte Carlo tree according to the space utilization rate, evaluating each iteration in the simulated boxing process, and finally taking a scheme formed by a series of optimal iterations as a boxing scheme;
in the process of obtaining a packing scheme according to the Monte Carlo tree search, recording the stacking relation of the cargos through an AABB tree, and judging whether each cargo meets the bearing limit;
the recording of the stacking relationship of the goods through the AABB tree and the judgment of whether each goods meets the bearing limit include:
when newly adding goods, obtaining a support object which is directly contacted with the bottom of the newly added goods;
updating the bearing information of each support according to the contact relationship between the weight of the newly added goods and the supports;
the updating of the bearing information of each support according to the weight of the newly added cargo and the contact relationship of the supports comprises:
connecting the newly added goods to the upper layer of the supports to obtain the contact area of the newly added goods and each support and the total contact area of the newly added goods and all the supports;
when the weight or the bearing information of the goods on the upper layer changes, the support on the layer updates the bearing information according to the change; and the number of the first and second electrodes,
the support of the layer distributes the change according to the ratio of the contact area of the support and the goods of the previous layer to the total contact area of the goods of the previous layer; wherein the content of the first and second substances,
the support at the lowest layer is a support surface of the container.
6. A boxing apparatus, wherein the boxing apparatus comprises a storage and a processor; the storage stores a boxing program;
the processor is configured to execute the boxing program to perform the steps of the boxing method of any one of claims 1 to 5.
7. A computer-readable storage medium, having stored thereon a binning program which, when executed by a processor, implements the steps of a binning method according to any of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811626116.0A CN109711624B (en) | 2018-12-28 | 2018-12-28 | Boxing method, equipment and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811626116.0A CN109711624B (en) | 2018-12-28 | 2018-12-28 | Boxing method, equipment and computer readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109711624A CN109711624A (en) | 2019-05-03 |
| CN109711624B true CN109711624B (en) | 2021-06-11 |
Family
ID=66259168
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811626116.0A Active CN109711624B (en) | 2018-12-28 | 2018-12-28 | Boxing method, equipment and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109711624B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111125878A (en) * | 2019-11-20 | 2020-05-08 | 深圳蓝胖子机器人有限公司 | Boxing method, system and computer readable storage medium |
| CN117114524B (en) * | 2023-10-23 | 2024-01-26 | 香港中文大学（深圳） | Logistics sorting method based on reinforcement learning and digital twin |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101381033A (en) * | 2008-10-15 | 2009-03-11 | 江苏科技大学 | Container loading method based on ant colony algorithm |
| CN101957945A (en) * | 2010-08-20 | 2011-01-26 | 上海电机学院 | Method and device for optimizing goods loading of container |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102841961B (en) * | 2012-07-20 | 2015-05-20 | 中南大学 | Method for detecting three-dimensional lifting dynamic collision based on cache information |
| CN104680237A (en) * | 2015-03-10 | 2015-06-03 | 西南科技大学 | Three-dimensional encasement novel genetic algorithm model under multi-constrain condition |
| CN105279629B (en) * | 2015-09-30 | 2018-09-21 | 中交第三航务工程勘察设计院有限公司 | A kind of intelligent packaging system of optimization |
| CN108015767B (en) * | 2017-11-30 | 2020-09-15 | 北京邮电大学 | Emergency operation method for space manipulator |
| CN108985458A (en) * | 2018-07-23 | 2018-12-11 | 东北大学 | A kind of double tree monte carlo search algorithms of sequential synchronous game |
-
2018
- 2018-12-28 CN CN201811626116.0A patent/CN109711624B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101381033A (en) * | 2008-10-15 | 2009-03-11 | 江苏科技大学 | Container loading method based on ant colony algorithm |
| CN101957945A (en) * | 2010-08-20 | 2011-01-26 | 上海电机学院 | Method and device for optimizing goods loading of container |
Non-Patent Citations (2)
| Title |
|---|
| Ranked Reward: Enabling Self-Play Reinforcement;Alexandre Laterre等;《arXiv：1807.01672v3》;20181206;第1-11页 * |
| 基于Web模式的3D装箱系统可视化关键技术及应用研究;王丽君;《中国优秀硕士学位论文全文数据库 信息科技辑》;20160215;第60-79页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109711624A (en) | 2019-05-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109711624B (en) | Boxing method, equipment and computer readable storage medium | |
| US11673743B2 (en) | Object handling coordination system and method of relocating a transporting vessel | |
| CN109670263B (en) | Boxing method, equipment and computer readable storage medium | |
| CN104408589B (en) | AGV Optimization Schedulings based on Hybrid Particle Swarm | |
| CN110443549B (en) | Method and system for generating boxing scheme of articles in logistics warehouse | |
| CN109665328B (en) | Dynamic optimization box stacking method and information data processing terminal | |
| CN109685278A (en) | Packing method, equipment and computer readable storage medium | |
| CN110175405B (en) | Vehicle loading optimization method and system | |
| JP7453722B2 (en) | Optimization method and system for multi-box three-dimensional loading based on multi-tree search | |
| CN109597304B (en) | Intelligent partitioned storage method for mold library based on artificial bee colony algorithm | |
| CN108510095B (en) | Method and device for determining goods picking path | |
| CN104915817A (en) | Loading method and apparatus for container | |
| CN112434893B (en) | Loading stacking type layer-by-layer optimal design method | |
| CN113222410B (en) | Method for establishing cargo space distribution model in bidirectional layout mode | |
| CN107330214A (en) | Spatial configuration optimal method based on discretization Yu heuristic evolution algorithm | |
| Klampfl et al. | Optimization of workcell layouts in a mixed-model assembly line environment | |
| CN109726841B (en) | AGV path calculation method based on unmanned cabin and AGV driving path control method | |
| CN108959782B (en) | Layout optimization method, device and equipment for intelligent workshop | |
| CN115629614B (en) | Storage multiple AGV path planning method based on genetic algorithm | |
| JP4025121B2 (en) | Component arrangement calculation device, component arrangement calculation method, component arrangement calculation program, recording medium recording the program, and component arrangement support system | |
| CN111125878A (en) | Boxing method, system and computer readable storage medium | |
| CN113495557A (en) | Method and device for determining number of target devices | |
| CN117252037B (en) | Three-dimensional boxing method and device, electronic equipment and storage medium | |
| Fujii et al. | Territorial and effective task decomposition for rearrangement planning of multiple objects by multiple mobile robots | |
| CN116777064B (en) | Two-dimensional boxing method based on non-primary cut constraint and branch pricing algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: B701-702, industrialization building, Shenzhen Virtual University Park, No.2, Yuexing Third Road, Nanshan District, Shenzhen, Guangdong Province Applicant after: Shenzhen Lan pangzi machine intelligence Co., Ltd Address before: B701-702, industrialization building, Shenzhen Virtual University Park, No.2, Yuexing Third Road, Nanshan District, Shenzhen, Guangdong Province Applicant before: SHENZHEN DORABOT ROBOTICS Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |