System and method for personalizing disassembly tasks distribution plan

Abstract

A method for personalizing a disassembly task distribution plan of a product between an operative and a robot in a human-robot collaborative cell, the method comprises processors to: detect structural features of the product to obtain disassembly modelling steps of the product 102 and joint nodes of the operator to obtain personalized user ergonomics data of the operator 104; obtain disassembly benefits and disassembly costs for disassembly tasks of the product 106; and provide a personalized disassembly tasks distribution plan of the product based on the disassembly modelling steps, the personalized user ergonomics data, the disassembly benefits and the disassembly costs of the product 108. The disassembly modelling steps of the product is done using an AND / OR graph (AOG). Personalized user ergonomics data is obtained using joint node angles of the operator from operator images. The plan is obtained using a list-based encoding and decoding scheme to model the disassembly tasks distribution plan for an optimizer.

Description

[0001] Various aspects of this disclosure relate to a system for personalizing disassembly tasks distribution plan. Various aspects of this disclosure relate to a method for personalizing disassembly tasks distribution plan. BACKGROUND

[0002] Disassembly plays a crucial role in smart manufacturing by enabling reverse engineering processes. Disassembly involves developing strategies and techniques for the efficient dismantling of products, with the aim of recovering valuable components, materials, and resources for reuse, remanufacturing, recycling, or proper disposal. Disassembly contributes to sustainability and circular economy by overturning the traditional "take-make-dispose" paradigm.

[0003] Most disassembly processes require operators to accomplish all tasks. Heavy workloads in repetitive disassembly significantly increase the risk of musculoskeletal disorders. The physical conditions vary among different operators, leading to ergonomic discrepancies when performing the same disassembly task. In comparison to a short operator, a taller operator needs to bend knees to a greater extent when performing tasks at the same height. Such ergonomics will increase the risk of musculoskeletal disorders and the potential disassembly costs.

[0004] By leveraging the disassembly capacities of both robots and humans, human-robot collaborative disassembly presents a promising opportunity for accommodating frequent changes resulting from diverse products. Therefore, personal disassembly ergonomics requires thorough attention, and a personalized disassembly planning approach should be developed by evaluating the capacity of robots and the ergonomics of operators with the aim of balancing disassembly cost and benefit.

[0005] However, it presents higher demands for disassembly planning to assign suitable tasks to suitable ones while making a good trade-off on disassembly profit and cost. Suitable ones may refer to a suitable candidate for disassembling the product. The suitable candidate may be an operator or a robot. Therefore, a multi-objective personalized disassembly planning approach for the human-robot collaborative cell by taking product structure, robot disassembly complexity, and human ergonomics into consideration simultaneously is desired. SUMMARY

[0006] Various embodiments concern a system for a personalizing disassembly tasks distribution plan of a product between an operative and a robot in a human-robot collaborative cell, the system including: one or more processor(s); and a memory having instructions stored therein, the instructions, when executed by the one or more processor(s), cause the one or more processor(s) to: detect structural features of the product to obtain disassembly modeling steps of the product; detect joint nodes of the operator to obtain personalized user ergonomics data of the operator; obtain disassembly benefits and disassembly costs for disassembly tasks of the product; provide a personalized disassembly tasks distribution plan of the product based on the disassembly modeling steps, the personalized user ergonomics data, the disassembly benefits and the disassembly costs of the product. Advantageously, by taking into account the personalized user ergonomics data, risk of musculoskeletal disorders of operators may decrease, while balancing disassembly cost and benefit.

[0007] According to one embodiment, the disassembly modeling steps of the product is done using a AND / OR graph (AOG), which indicates how subcomponents interact with other subcomponents in the product.

[0008] According to one embodiment, the disassembly modeling steps is based on the structural features of the product and disassembly constraints of the product.

[0009] According to one embodiment, the personalized user ergonomics data of the operator is obtained using joint node angles of the operator.

[0010] According to one embodiment, the disassembly costs comprise complexity of the robot in performing disassembly tasks and the personalized user ergonomics data of the operator.

[0011] According to one embodiment, the one or more processor(s) is further configured to calculate an operation time of disassembling the product by the operator and the robot using the disassembly modeling steps, and the personalized user ergonomics data.

[0012] According to one embodiment, the personalized disassembly tasks distribution plan is obtained using a list-based encoding and decoding scheme based on the operation time, the disassembly benefits and the disassembly costs of the product.

[0013] According to one embodiment, the list-based encoding and decoding scheme is used to model the disassembly tasks distribution plan for an optimizer.

[0014] According to one embodiment, the personalized disassembly tasks distribution plan differs between different operators for a same product. Since the physical conditions vary among different operators, leading to discrepancies in ergonomics when performing the same disassembly task. Such ergonomics significantly increase the risk of injury. Having different personalized disassembly tasks distribution plan for each operator will reduce the risk of injury.

[0015] Various embodiments concern a method for personalizing a disassembly tasks distribution plan of a product between an operative and a robot in a human-robot collaborative cell, the method comprising using one or more processor(s) to: detect structural features of the product to obtain disassembly modeling steps of the product; detect joint nodes of the operator to obtain personalized user ergonomics data of the operator; obtain disassembly benefits and disassembly costs for disassembly tasks of the product; provide a personalized disassembly tasks distribution plan of the product based on the disassembly modeling steps, the personalized user ergonomics data, the disassembly benefits and the disassembly costs of the product.

[0016] According to one embodiment, the method further comprising: using a AND / OR graph (AOG) to obtain the disassembly modeling steps of the product.

[0017] According to one embodiment, the disassembly modeling steps is based on the structural features of the product and disassembly constraints of the product.

[0018] According to one embodiment, the method further comprising: obtaining the personalized user ergonomics data of the operator using joint node angles of the operator.

[0019] According to one embodiment, the disassembly benefits comprise recycling revenues of subcomponents of the product.

[0020] According to one embodiment, the disassembly costs comprise complexity of the robot in performing disassembly tasks and the personalized user ergonomics data of the operator.

[0021] According to one embodiment, the method further comprising: calculating an operation time of disassembling the product by the operator and the robot using the disassembly modeling steps, and the personalized user ergonomics data.

[0022] According to one embodiment, the method further comprising: obtaining the personalized disassembly tasks distribution plan using a list-based encoding and decoding scheme based on the operation time, the disassembly benefits and the disassembly costs of the product.

[0023] According to one embodiment, the list-based encoding and decoding scheme is used to model the disassembly tasks distribution plan for an optimizer.

[0024] According to one embodiment, the personalized disassembly tasks distribution plan differs between different operators for a same product.

[0025] It should be noted that embodiments described in context of the method for personalizing disassembly tasks distribution plan are analogously valid for the system for personalizing disassembly tasks distribution plan and vice versa. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The present disclosure will be better understood with reference to the detailed description when considered in conjunction with the non-limiting examples and the accompanying drawings, in which: - FIG. 1 illustrates a method for personalizing disassembly tasks distribution plan according to various embodiments. - FIG. 2 illustrates a system for personalizing disassembly tasks distribution plan according to various embodiments. - FIG. 3 illustrates an exemplary disassembly task modelling of a product according to various embodiments. - FIGS. 4A and 4B illustrate an exemplary detection of personalized user ergonomics data according to various embodiments. - FIGS. 5A and 5B illustrates exemplary encoding and decoding scheme according to various embodiments. - FIG. 5C shows an example of discrete random walks. - FIGS. 6A and 6B illustrates exemplary personalized disassembly tasks distribution plan for different operators according to various embodiments. DETAILED DESCRIPTION

[0027] The following detailed description refers to the accompanying drawings that show, by way of illustration, specific details and embodiments in which the disclosure may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the disclosure. Other embodiments may be utilized and structural, and logical changes may be made without departing from the scope of the disclosure. The various embodiments are not necessarily mutually exclusive, as some embodiments can be combined with one or more other embodiments to form new embodiments.

[0028] Embodiments described in the context of one of the devices or methods arc analogously valid for the other devices or methods. Similarly, embodiments described in the context of a device are analogously valid for a method, and vice-versa.

[0029] Features that are described in the context of an embodiment may correspondingly be applicable to the same or similar features in the other embodiments. Features that are described in the context of an embodiment may correspondingly be applicable to the other embodiments, even if not explicitly described in these other embodiments. Furthermore, additions and / or combinations and / or alternatives as described for a feature in the context of an embodiment may correspondingly be applicable to the same or similar feature in the other embodiments.

[0030] In the context of various embodiments, the articles “a”, “an” and “the” as used with regard to a feature or element include a reference to one or more of the features or elements.

[0031] As used herein, the term “and / or” includes any and all combinations of one or more of the associated listed items.

[0032] FIG. 1 illustrates a method for personalizing disassembly tasks distribution plan according to various embodiments.

[0033] According to various embodiments, the method 100 for personalizing disassembly tasks distribution plan of a product between an operative and a robot in a human-robot collaborative cell may be provided. In some embodiments, the method 100 may include a step 102 of using one or more processor(s) to detect structural features of the product to obtain disassembly modeling steps of the product. The products may be lithium batteries, household appliances, vehicle control boxes or any suitable product.

[0034] In various embodiments, personalizing disassembly tasks distribution plan may mean the order of which components of a product are disassembled and whether the disassembly of the component is assigned to an operative or a robot.

[0035] In various embodiments, the method 100 may also include a step 104 of detecting joint nodes of the operator to obtain personalized user ergonomics data of the operator.

[0036] In various embodiments, the method 100 may also include a step 106 of obtaining disassembly benefits and disassembly costs for disassembly tasks of the product.

[0037] In various embodiments, the method 100 may also include a step 108 of provide a personalized disassembly tasks distribution plan of the product based on the disassembly modeling steps, the personalized user ergonomics data, the disassembly benefits and the disassembly costs of the product.

[0038] Steps 102 to 108 are shown in a specific order, however other suitable arrangements are possible. Steps may also be combined in some cases. Any suitable order of steps 102 to 108 may be used.

[0039] FIG. 2 illustrates a system for personalizing disassembly tasks distribution plan according to various embodiments.

[0040] In an embodiment, the system 200 may include at least one of: a processor 202 for personalizing disassembly tasks distribution plan, a memory 204, an input / output module 206, and a computer vision module 208 which may include a vision detection device such as a camera.

[0041] While such program modules are shown in block diagram format, it is to be understood that the functionality performed by a single program module or by multiple program modules.

[0042] In an embodiment, the memory 204 may be a random-access memory (RAM). The memory 204 may be used to store instructions for personalizing disassembly tasks distribution plan.

[0043] According to various embodiments, a computer program product may store the computer executable code including instructions for personalizing disassembly tasks distribution plan according to the various embodiments. The computer executable code may be a computer program. The computer program product may be a non-transitory computer-readable medium. The computer program product may be in or may be the system 200.

[0044] Various embodiments concern a system for personalizing disassembly tasks distribution plan of a product between an operative and a robot in a human-robot collaborative cell is disclosed.

[0045] In various embodiments, the processor 202 may detect structural features of the product to obtain disassembly modeling steps of the product. The processor 202 may receive information regarding the structural features of the product from the computer vision module 208 which may include an image captuxing device, e.g. a camera for capturing the image of the product. The processor 202 may detect or deduce the structural features of the product from the image of the product provided by the camera.

[0046] In various embodiments, from the structural features, the processor 202 may obtain disassembly modeling steps of the product. The disassembly modeling steps of the product may include instructions or directions on which components should be removed before another component.

[0047] In various embodiments, the product may be lithium batteries, household appliances, vehicle control boxes or any suitable product.

[0048] In various embodiments, the disassembly modeling steps of the product may be done using a AND / OR graph (AOG). An AOG is a graphical representation of the reduction of problems to conjunctions and disjunctions of subproblems. The AOG can be derived based on analyzing the design / manufacturing documents of the product, e.g., CAD models and machining information. Such documents indicate the components / subcomponents and interrelationships that exist in the product. By analyzing this information, engineers can construct the AOG based on the precedent and exclusive relationship of subcomponents.

[0049] In various embodiments, the disassembly modeling steps is based on the structural features of the product and disassembly constraints of the product. The disassembly constraints may include directions on which component needs to be removed before another component. For example, the product may include components A, B, C, wherein component B may only be removed after component A is removed. This may be due to the location of the various components in the product. For example, component B may be located below component A, therefore, component A needs to be removed before component B can be removed.

[0050] In various embodiments, the processor 202 may detect joint nodes of the operator to obtain personalized user ergonomics data of the operator. The processor 202 may receive information regarding the joint nodes from the computer vision module 208 which may include a camera for capturing the image of the operator. The processor 202 may detect or deduce the joint nodes from the image of the operator provided by the camera. The processor 202 may detect or deduce the joint nodes through a computer vision algorithm, AI algorithm or deep learning methods.

[0051] In various embodiments, the personalized user ergonomics data of the operator is obtained using joint node angles of the operator. The joint node angles may be or may include the degree that the operator’s knees, elbows or any other relevant joints are bent while removing components from the products.

[0052] In various embodiments, the processor 202 may obtain disassembly benefits and disassembly costs for disassembly tasks of the product.

[0053] In various embodiments, disassembly benefits may be the benefits or advantages from disassembly components in a product. In various embodiments, the disassembly benefits include recovering valuable components, materials, and resources for reuse, remanufacturing, recycling, proper disposal or recycling revenues of subcomponents of the product.

[0054] In various embodiments, disassembly costs may be the costs or disadvantages from disassembly components in a product. In various embodiments, the disassembly costs includes complexity of the robot in performing disassembly tasks and the personalized user ergonomics data of the operator. In various embodiments, disassembly costs may be the costs or disadvantages of disassembly operations. In various embodiments, the disassembly costs include both the complexity of the robot and the ergonomics of the operator in performing disassembly tasks.

[0055] In various embodiments, the processor may take into account both the disassembly benefits and disassembly costs, when deciding whether a robot or operator should disassemble a component from the product.

[0056] In various embodiments, the processor 202 may provide a personalized disassembly tasks distribution plan of the product based on the disassembly modeling steps, the personalized user ergonomics data, the disassembly benefits and the disassembly costs of the product.

[0057] In various embodiments, the processor 202 may calculate an operation time of disassembling the product by the operator and the robot using the disassembly modeling steps, and the personalized user ergonomics data.

[0058] In various embodiments, the personalized disassembly tasks distribution plan may be obtained using a list-based encoding and decoding scheme based on the operation time, the disassembly benefits and the disassembly costs of the product.

[0059] In various embodiments, the list-based encoding and decoding scheme may be used to model a disassembly plan for an optimizer. The optimizer may be a hybrid multi-objective ant lion optimizer (HMALO), or any other suitable optimizer. An ant lion optimizer is a metaheuristic that mathematically models the interaction of ants and antlions in nature. The optimizer solves optimization problems considering random walk of ants, building traps, entrapment of ants in traps, catching preys, and re-building traps.

[0060] According to one embodiment, the personalized disassembly tasks distribution plan differs between different operators for a same product. For example, for a product which includes components A, B and C, for operator A, the operator A may be instructed to remove components A and B while the robot removes component C but for operator B, the operator B may be instructed to remove components A while the robot removes components B and C. This may be because the physical conditions vary among different operators, leading to ergonomic discrepancies when performing the same disassembly task. In comparison to a short operator, a taller operator needs to bend knees to a greater extent when performing tasks at the same height. Therefore, when calculating the disassembly benefits and the disassembly costs of the product, the personalized disassembly tasks distribution plan may differ between different operators for a same product.

[0061] FIG. 3 illustrates an exemplary disassembly task modelling of a product according to various embodiments.

[0062] In various embodiments, the processor may detect structural features of a product 300 to obtain disassembly modeling steps of the product. The processor may receive information regarding the structural features of the product 300 from a camera for capturing the image of the product 300.

[0063] In an example, the product 300 may include components 302. The components 302 may include 9 components, the 9 components comprising components A to I.

[0064] In various embodiments, from the structural features, the processor may obtain disassembly modeling steps of the product. The disassembly modeling steps of the product may include instructions or directions on which components should be removed before another component.

[0065] In various embodiments, the disassembly modeling steps of the product 300 may be done using a AND / OR graph (AOG) 304.

[0066] AOG-based modeling depicts the subcomponents of a product together with alternative disassembly tasks according to logical combination. Taking subcomponents as nodes and disassembly tasks as arcs (e.g., hyper-arcs), the AOG of the product 300 can be constructed as FIG. 3. Each subcomponent is represented by a rectangular node, which indicates its index and constituent elements. Every hyper-arc links two nodes by AND relation to reveal a disassembly task. Moreover, OR relation indicates that a node has various decompositions, but only one of them can be selected in a disassembly task plan.

[0067] According to AOG 304, disassembly tasks can be revealed by AND relations and mathematically expressed by a transition matrix ' ~ i -I if 1 tass J OJ where N is the number of disassembly tasks and AY is the number of subcomponents. Moreover, disassembly precedence is represented by a precedence matrix ’ x i X'-' A-■ In order to depict OR relations, an exclusive matrix ' S! "? 5 is defined to ensure that only one disassembly task is executed for a subcomponent: f~; sf is A s i »4 / are «; xs Of: sskti-xsi * i &ofesxme

[0068] In various embodiments, the disassembly modeling steps is based on the structural features of the product and disassembly constraints of the product 300. The disassembly constraints may include directions on which components 302 needs to be removed before another component. For example, the product may include components A to I, wherein the first step may be to removed component A from components BCDEFGHI. The second step may be to remove component D from components BCEFGHI. This may be due to the location of the various components in the product. For example, component A may be located or screwed above the other components, therefore, component A needs to be removed before the other components can be removed.

[0069] FIGS. 4A and 4B illustrate an exemplary detection of personalized user ergonomics data according to various embodiments.

[0070] Bad ergonomics in repetitive work is one of the causes of work-related musculoskeletal disorders (MSD). By integrating ergonomic principles into disassembly task planning, a human-robot hybrid cell can be used to effectively prioritize the prevention of MSD. However, the physical conditions vary among different operators, leading to discrepancies in ergonomics when performing the same disassembly task. For instance, in comparison to a short operator, a taller one needs to bend knees to a greater extent when performing tasks at the same height. Such ergonomics significantly increase the risk of MSD.

[0071] In various embodiments, the processor may detect joint nodes of the operator to obtain personalized user ergonomics data of the operator. The processor may receive information regarding the joint nodes from a camera for captuxing the image of the operator. The processor may detect or deduce the joint nodes from the image of the operator provided by the camera. The camera may be a low-cost vision system as long as the angles of the operator’s joints can be captured.

[0072] In various embodiments, the personalized user ergonomics data of the operator is obtained using joint node angles of the operator. The joint node angles may be or may include the degree that the operator’s knees, elbows or any other relevant joints are bent while removing components from the products.

[0073] For example, in the example shown in FIG. 4A, the operator’s trunk is bent at an angle of 30 degrees while removing the component from the product. According to the Rapid entire body assessment (REBA) principle, work-related musculoskeletal disorders can be assessed automatically based on the body angles measured by the system.

[0074] REBA is a method for evaluating MSD based on ergonomics. Main body postures, e.g. trunk, legs, and arms, are analyzed and corresponding scores are determined. However, it generally requires an expert to observe and manually annotate postures, which is labor-intensive. As REBA scores are mainly determined based on body angles, a vision system may be used to detect the joint nodes of an operator. The vision system may be implemented using a vision device (e.g., a camera) and a vision algorithm such as an open source vision algorithm by MediaPipe which may reconstruct a 3D body pose from a 2D image of an operator.

[0075] As shown in FIG. 4A, a skeleton structure of an operator is detected based on a 2D image from the vision device, and a 3D body pose may be reconstructed from the 2D image using the vision algorithm. According to the coordination of joint nodes, the body angles can be estimated and its corresponding score is decided. The comprehensive REBA value for a disassembly task is derived by coupling the scores as shown in FIG. 4B. It is deemed as a key parameter representing the MSD risk of different operators and is considered in a mathematical model for personalized disassembly task planning.

[0076] FIGS. 5A and 5B illustrates exemplary encoding and decoding scheme according to various embodiments.

[0077] Personalized features of an operator yield different disassembly costs. In order to make a tradeoff between disassembly benefits and costs, a multi-objective personalized disassembly sequence planning (p-DSP) model is developed for the human-robot collaborative cell. The recycling revenues of subcomponents are calculated as disassembly benefits whereas the disassembly complexity of robots and ergonomics of the operators are deemed as disassembly costs. The operator’s personal working experiences are reflected by their working time.

[0078] Disassembly complexity is a valuable indicator of how suitable a robot is for performing a disassembly task. It is defined by eight criteria listed in Table I. Each criterion possesses varying levels of complexity, which are assigned corresponding scores. is* is :• -4 sssss$&ssii>- Z .V i Low X> > v 4.5 Samstag ? A Mataw Wt MtMT J&sSs su-ow;- 5.$ 7 A

[0079] The disassembly complexity is calculated as the cost of a robot in executing the disassembly task i: lx where 'is the evaluation results regarding the m-th parameter' in Table I. If a disassembly taskj is out of a robot’s capacity, then a ’

[0080] Taking task execution parameter di, disassembly sequence pair Sji, and task allocation parameters ri and h; as decision variables, two objective functions are represented as follows: In various embodiments, b is the recycling values of subcomponents, X is the the relationship between subcomponents and their corresponding disassembly tasks.. serves as a quantitative assessment of the cost paid by the operator according to ergonomics. tc is the transaction time between the operator / robot role changing.

[0081] Mtn T - VC k (2

[0082]

[0083] where B aims to disassemble high-profit subcomponents with low operator’s and robot’s working costs. T aims to minimize operating times in a human-robot collaborative cell, f f bk is the recycling values of subcomponents, >and ' are the working time of a robot and operator on executing disassembly task i, respectively. Constraints can be represented according to the AOG and task allocation rules: s: ~ 0, VO, ~ os COsMV * {0.1

[0084] '" ' ' t

[0085] In various embodiments, constraint (3) ensures that each disassembly task is performed only once.

[0086] In various embodiments, constraint (4) reveals that at least one disassembly task is selected and the precedent relations should be satisfied in a disassembly sequence. The exclusive constraints are ensured in constraint (5).

[0087] In various embodiments, constraint (6) indicates the equilibrium of the in-degree and out-degree of a disassembly task.

[0088] In various embodiments, constraints (7) and (8) state that a selected disassembly task can be assigned to either a robot or an operator.

[0089] In various embodiments, constraint (9) ensures that a disassembly is not assigned to the robot if it is out of its capacity.

[0090] In various embodiments, constraint (10) limits all the decision variables to be binary.

[0091] In order to solve the p-DSP model for the human-robot collaborative disassembly, a metaheuristic optimizer, namely hybrid multi-objective ant lion optimizer (HMALO), is developed with quadruple list-based encoding / decoding schemes and hybrid updating strategies. The obtained Pareto result can be deemed as an optimal disassembly plan for different operators in the human-robot collaborative cell. The encoding scheme which includes variables such as the operation time, the disassembly benefits and the disassembly costs of the product is an input generated for the ant lion optimizer, based on which the ant lion optimizer can acquire the optimal disassembly plans. Further, the optimal disassembly plans from the optimizer will be decoded to generate the personalized disassembly tasks distribution plan for each different operator.

[0092] An ant lion optimizer is a meta-heuristic that mathematically models the interaction of ants and antlions in nature. The ant lion optimizer solves optimization problems considering steps as such random walk of ants, discrete random walk updating, binary random walk updating and catching prey and archive saving.

[0093] Ants and antlions are encoded to represent potential disassembly task plans. Hybrid features are accounted for since disassembly tasks should be encoded by discrete values whereas task assignments can be encoded as binary ones. A quadruple list-based hybrid encoding scheme is used to simultaneously reveal both disassembly sequence and task allocation. As shown in FIG. 5A, a quadruple list includes two parts namely sequence and allocation. The sequence indicates the discrete permutation of all disassembly tasks in the first list. It decides whether disassembly tasks are executed in the second list by using binary values. The allocation part determines whether a selected task is assigned to a robot or operator. The encoded quadruple list can be deemed as the position of ants or antlions in the hyperspace.

[0094] Antlions prey on ants and then change positions to improve the chance of catching new ants. The positions of antlions after iteration are decoded as disassembly task plans. For instance, the quadruple list of the antlion in FIG. 5B reveals that disassembly tasks 1-2-4-6-8 are executed in sequence. As shown in FIG. 5B, different tasks are assigned to the robot and operator.

[0095] The robot undertakes tasks 1,2, and 6 whereas the operator is responsible for tasks 4 and 8.

[0096] Disassembly precedence and exclusion are two types of constraints that should be satisfied to ensure the feasibility of disassembly task plans. Adjusting should be made for the quadruple list to satisfy all constraints in the proposed mathematical model. Exclusive constraints arise from two aspects namely OR relations defined in AOG, and exclusive task assignment. Taking the quadruple list and exclusive matrix as input, the disassembly tasks in the sequence part are checked first for removing exclusive tasks from the left to right side. If a disassembly task is chosen, it should be assigned to either a robot or operator. The exclusion adjustment is realized in Algorithm 1. Algorithm!: adjusting exclusive disassembly tasks______________ Input: The quadruple encoded list, the exclusive matrix E Output: New snt'asitiions represented by a quadruple list For Every executed disassembly task i 'in the sequence part For Every executed disassembly ta.sk / at the right side off If i and / are exclusive tasks shown in E Set the variable in the second list to 0 End End IF 'The sum of variable in the third and fourth list != 1 Randomly assign i-ik disassembly task to either a robot or operator End End For Every nan-executed disassembly A in the sequence part Set the £-th variable in die third and fourth lists to 0 End Precedence constraints consider the structure of a product and ensure that precedent components have been disassembled. On one hand, the permutation of the quadruple list should

[0097]

[0098] be adjusted to satisfy precedence constraints. On the other hand, necessary disassembly tasks may be added as immediate predecessors. An adjusting strategy first determines whether an immediate predecessor of the current disassembly task has been executed. If not, the permutation of it is adjusted in the quadruple list. The strategy for satisfying precedence constraints is realized in Algorithm 2. Algon;h»» 2.- adjasftBg ths g of disassgaihty tasks_______ laput: The encoded list., the precedence matrix. F rhe exclusive matrix £ Output: New at&MXt&Kn rejseseaied by aqasdatple list a « Number of the executed dwas-settFlv tasks WA a >0 Csfeuteie she mmitmtm isxtmediate predecessor aassbers g of all executed rhsassemb'iv tasks based us F If q = = 0 Find the srtdex r of the task with no iramedsste predecessor a ::;n 1 Efe Rad a new task set without immediate ps-edecessws according to ths updated F For Every msw wk / If J is exclusive tn the executed tasks Delete.» &»ra the sew task sei End End If The amber of variables ia die new task set >$ Choos; a new task and find ds index s End Set she i-ih variable m the second hst to 1 Raixkstdy assign j -th task to either a rate us an operator End Change the premutation of t -th. variable j» the qwdnspfe list Update Fto exclude Ms dtsassesnW task End

[0099] -===-------------------------------------------------

[00100] The general steps of the ant lion optimizer is described below.

[00101] Random walk of ants is the part of the antlion optimizer that ensures exploration and exploitation. This part reveals the random search abilities of ants in the hyperspace, which is influenced by the trapping, sliding, preying, and pits re-constructing behaviors of antlions. The initial position of an ant in dimension d after random walks can be represented as:

[00102] ' fe"' v 5 ? jfusucdj

[00103] where t is the current iteration step, is the function that generates a random walk and rand is a random number in [0, 1]. The entrapment of ants in the pits constructed by antlions affects the rand walk. The range of the pits can be defined by: D: ~ C* 1 + c' ,. -- CO ¢7 ?' '

[00104] where ‘ ; are vectors that indicate the minimum and maximum positions of the i-th ant at the t-th iteration, respectively. reveals the position of a selected antlion. c and are the lower and upper limits of position changes at t-th iteration. Once ants are in the trap, antlions shoot sand from the pits to slide down ants for preying. It can be simulated by the decreased range of position changes: / . \ / / 11 I i 1 - 7 i \ 4 / I T) where ; is the maximal iteration count, w is a coefficient determined by the current iteration step aiming to adjust the accuracy level of exploitation. Taking the hunting behaviors of antlions into consideration, the final random walk J : can be adjusted as: A.....:.............11.,..3.....:.............:..14. (r s (x-X)

[00105] where is the initial random walks of the z'-th ant at the i-th iteration, and A ■<Uia *3 are the minimum and maximum values of the z-th ant’s initial random walk.

[00106] In a discrete antlion optimizer, a reversing operator is utilized instead of random walks. Since the updating of discrete disassembly tasks should not add, repeat or delete existing ones, the discrete random walk can be simulated as the re-permutation of the quadruple list, ( / )))< i e, "■ - ■ v ,- \ ' "■ .• /

[00107] where Xs thc / -th variable in the first list that represents a disassembly task. 'X ’ 1 ( / ) x v) round the random walk ' ' and ' moves the disassembly taskx withy positions. FIG. 5C shows an example of discrete random walks. In FIG. 5C, a positive value of y results in right moving, while a negative one lead to left moving.

[00108] In the step of binary random walk updating, the execution of disassembly tasks and their allocations are represented by binary values in the quadruple list. The binary updating approach is investigated based on random walks. Since random walks are in the form of continuous values, a transfer function is applied to convert them into 0 or 1, which makes ants search in a binary hyperspace: n' I / 1

[00109] where ' ' is the random walk of the j-th variable regarding a binary list of the i-th ant. ’ ' ' reveals the updating probability, which is further converted to a binary value based on a random threshold: H if if ( / ) Strand [0 otherwise where f' is the updated binary value of thej-th dimension in the i-th ant for iteration t+1. The operator can be applied to handle the effect of different antlions on the random walk of ants.

[00110] In the step of catching prey and archive saving, ants search the hyperspace for nondominated disassembly plans. Once an ant dominates an antlion during iterations, the antlion preys on it and then changes to a new place for further preying. Such behaviors are represented as: where reveals the dominance. Thus, antlions are deemed as Pareto optimal solutions and stored in an archive during iterations. The leader selection and archive maintenance methods are utilized. The selected antlion is regarded as an elite that impacts the random walk of ants.

[00111] FIGS. 6A and 6B illustrates exemplary personalized disassembly tasks distribution plan for different operators according to various embodiments.

[00112] As shown in FIGS. 6A and 6B, for operator 1, the robot is assigned to execute much more disassembly tasks. This is because operator 1 exhibits high musculoskeletal disorder risks, which results in higher disassembly costs than the robot. On the contrary, operator 2 is tasked with more disassembly tasks than the robot. This is influenced by the experience and relatively good ergonomics of operator 2, which make her more cost-effective than the robot in disassembly. In general, the developed technique reduces the workload of personal operators while balancing the disassembly cost and benefit of the manufacturing company.

[00113] The methods described herein may be performed and the various processing or computation units and the devices and computing entities described herein may be implemented by one or more circuits. In an embodiment, a "circuit" may be understood as any kind of a logic implementing entity, which may be hardware, software, firmware, or any combination thereof. Thus, in an embodiment, a "circuit" may be a hard-wired logic circuit or a programmable logic circuit such as a programmable processor, e.g. a microprocessor. A "circuit" may also be software being implemented or executed by a processor, e.g. any kind of computer program, e.g. a computer program using a virtual machine code. Any other kind of implementation of the respective functions which are described herein may also be understood as a "circuit" in accordance with an alternative embodiment.

[00114] While the disclosure has been particularly shown and described with reference to specific embodiments, it should be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the present disclosure as defined by the appended claims. The scope of the present disclosure is thus indicated by the appended claims and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced.

Claims

1. A system for personalizing a disassembly tasks distribution plan of a product between an operative and a robot in a human-robot collaborative cell, the system comprising:one or more processor(s); anda memory having instructions stored therein, the instructions, when executed by the one or more processor(s), cause the one or more processor(s) to:detect structural features of the product to obtain disassembly modeling steps of the product;detect joint nodes of the operator- to obtain personalized user ergonomics data of the operator;obtain disassembly benefits and disassembly costs for disassembly tasks of the product;provide a personalized disassembly tasks distribution plan of the product based on the disassembly modeling steps, the personalized user ergonomics data, the disassembly benefits and the disassembly costs of the product.

2. The system of claim 1, wherein the disassembly modeling steps of the product is done using a AND / OR graph (AOG).

3. The system of claim 1 or claim 2, wherein the disassembly modeling steps is based on the structural features of the product and disassembly constraints of the product.

4. The system of any one of claims 1-3, wherein the personalized user ergonomics data of the operator is obtained using joint node angles of the operator.

5. The system of any one of claims 1-4, wherein the disassembly benefits comprise recycling revenues of subcomponents of the product.

6. The system of any one of claims 1-5, wherein the disassembly costs comprise complexity of the robot in performing disassembly tasks and the personalized user ergonomics data of the operator.

7. The system of any one of claims 1-6, wherein the one or more processor(s) is further configured to calculate an operation time of disassembling the product by the operator and the robot using the disassembly modeling steps, and the personalized user ergonomics data.

8. The system of claim 7, wherein the personalized disassembly tasks distribution plan is obtained using a list-based encoding and decoding scheme based on the operation time, the disassembly benefits and the disassembly costs of the product.

9. The system of claim 8, wherein the list-based encoding and decoding scheme is used to model the disassembly tasks distribution plan for an optimizer.

10. The system of any one of claims 1-9, wherein the personalized disassembly tasks distribution plan differs between different operators for a same product.

11. A method for personalizing a disassembly tasks distribution plan of a product between an operative and a robot in a human-robot collaborative cell, the method comprising using one or more processor(s) to:detect structural features of the product to obtain disassembly modeling steps of the product;detect joint nodes of the operator to obtain personalized user ergonomics data of the operator;obtain disassembly benefits and disassembly costs for disassembly tasks of the product;provide a personalized disassembly tasks distribution plan of the product based on the disassembly modeling steps, the personalized user ergonomics data, the disassembly benefits and the disassembly costs of the product.

12. The method of claim 11, the method further comprising:using a AND / OR graph (AOG) to obtain the disassembly modeling steps of the product.

13. The method of claim 11 or claim 12, wherein the disassembly modeling steps is based on the structural features of the product and disassembly constraints of the product.

14. The method of any one of claims 11-13, the method further comprising: obtaining the personalized user' ergonomics data of the operator using joint node angles of the operator.

15. The method of any one of claims 11-14, wherein the disassembly benefits comprise recycling revenues of subcomponents of the product.

16. The method of any one of claims 11-15, wherein the disassembly costs comprise complexity of the robot in performing disassembly tasks and the personalized user ergonomics data of the operator.

17. The method of any one of claims 11-16, the method further comprising: calculating an operation time of disassembling the product by the operator and the robot using the disassembly modeling steps, and the personalized user ergonomics data.

18. The method of claim 17, the method further comprising:obtaining the personalized disassembly tasks distribution plan using a list-based encoding and decoding scheme based on the operation time, the disassembly benefits and the disassembly costs of the product.

19. The method of claim 18, wherein the list-based encoding and decoding scheme is used to model the disassembly tasks distribution plan for an optimizer.

20. The method of any one of claims 11-19, wherein the personalized disassembly tasks distribution plan differs between different operators for a same product.

Images