Disassembly-based Assembly Planning

Abstract

Robot motion planning technology for component assembly operations. The input to the motion planning technology includes the geometric models of the individual components to be assembled and the initial and target arrangements. The method begins with a tightly constrained goal or final arrangement and plans in the direction of the loosely constrained initial arrangement. A randomly sampled waypoint arrangement is proposed, followed by a local search for a feasible arrangement that generates multiple nodes that extend paths towards the initial arrangement while sliding through the tightly constrained regions. For a given randomly sampled arrangement, the local search can be repeated multiple times. When a complete path is found, the motion sequence is trimmed to remove unnecessary extraneous motions in the loosely constrained regions. The disclosed method significantly reduces the number of poorly evaluated arrangements and finds assembly solutions faster compared to known tree-based motion planning methods.

Description

[0001] Cross - Reference to Related Applications

[0002] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 010,724, filed on April 16, 2020, titled "Disassembly-Based Assembly Planning". Technical Field

[0003] The present disclosure relates to the field of industrial robot motion planning and, more particularly, to robot motion planning techniques for component assembly operations that begin with a tightly constrained goal configuration and plan in the direction of a loosely constrained initial configuration, where local search follows randomly sampled configurations, generating multiple nodes that extend paths toward the initial configuration while sliding through the tightly constrained regions. Background Art

[0004] It is well known to use industrial robots to perform a variety of manufacturing, assembly, and material handling operations. One such application is using one or more robots to assemble individual components into a product. In component assembly operations, it is very common for the assembly configuration of two components to include multiple contact points and surfaces such as a protrusion in a slot, a pin in a hole, direct contact of adjacent surfaces, etc., in order to hold the two components in a proper relative orientation. These contact points and surfaces represent constraints that require very precise assembly motion when the two components are brought together.

[0005] People are good at assembling components in the above manner using vision, touch, and intuition. However, for reasons such as cost-effectiveness and consistent quality, it is desirable to use robots to perform such repetitive tasks. Using humans to teach robots the correct assembly sequence is time-consuming and expensive because an accurate motion sensing environment must be established to achieve accurate motion capture of the assembly sequence. In addition, human assembly of components typically involves many trial-and-error processes, characterized by small movements to fit the components into their correct engagement, and those small irrelevant movements are not desired to be captured in the robot motion program. For these reasons, computer-based assembly motion planning techniques have been evaluated to replace human teaching.

[0006] The prior art of computer-based assembly motion planning has been disclosed, which is based on the Rapidly-exploring Random Tree (RRT) method. The RRT quickly and randomly builds a space-filling tree of nodes until it finds a path from the initial configuration to the final configuration. Given the mathematical models (surface or solid models) of two parts to be assembled and their initial and final configurations relative to each other, the RRT can be used to find the assembly motion sequence. However, due to the tightly constrained target configuration of the two components, the chance for the RRT to find a path waypoint that defines a motion sequence with the correct relative position and orientation of the parts exactly is very low. Therefore, RRT-based techniques for assembly planning of tightly constrained parts have been found to be inefficient.

[0007] Other known computer-based assembly planning systems are rule-based and use geometric constraints and rules, such as planar and cylindrical constraints on the relative component configurations. This type of system can be more effective than RRT-based techniques, but usually can only handle simple part geometries due to the need to define rules for each potential interface.

[0008] In view of the above, there is a need for an improved robotic motion planning technique for part assembly operations that can handle the complex geometries typical of real parts and can also effectively plan the assembly motion for tightly constrained parts. Summary of the Invention

[0009] In accordance with the teachings of the present disclosure, a robotic motion planning technique for part assembly operations is disclosed. The input to the motion planning technique is the geometric models of the parts to be assembled and the initial and target configurations. The method starts from the tightly constrained target or final configuration and plans in the direction of the loosely constrained initial configuration. Randomly sampled waypoints are proposed, followed by a local search for feasible configurations, which generates a number of nodes that extend towards the initial configuration while sliding through the path of the tightly constrained region. The local search can be repeated multiple times for a given randomly sampled configuration. When a complete path is found, the motion sequence is trimmed to remove unnecessary extraneous motions in the loosely constrained region. The disclosed method significantly reduces the number of poorly evaluated configurations and finds the assembly solution faster compared to known tree-based motion planning methods.

[0010] In conjunction with the accompanying drawings, additional features of the presently disclosed apparatus and methods will become apparent from the following description and the appended claims. Brief Description of the Drawings

[0011] Figure 1A 、 1B Figures 1C and 1D are schematic views of two parts being assembled, where traditional robotic motion planning techniques are very inefficient and highlight the need for improved motion planning techniques;

[0012] Figure 2 is a simple assembled two-dimensional schematic of a ball entering a slot, showing the loose constraint nature of the initial arrangement and the tight constraint nature of the target arrangement, as typically encountered in component assembly operations;

[0013] Figure 3 is according to techniques known in the art Figure 2 of a ball and slot assembly layout, where a random tree motion planning method has difficulty finding a path into the tightly constrained region;

[0014] Figure 4 is according to an embodiment of the present disclosure Figures 2 - 3 of a ball and slot assembly layout, where motion planning starts from the target arrangement, selects a randomly chosen waypoint arrangement, and defines a plurality of local samples;

[0015] Figure 5 is according to an embodiment of the present disclosure Figure 4 of a ball and slot assembly layout, where the local sample arrangements are evaluated to identify which sample is collision-free and closest to the randomly chosen arrangement;

[0016] Figure 6 is according to an embodiment of the present disclosure Figures 4 - 5 of a ball and slot assembly layout, where additional local sample arrangements are evaluated relative to the same previously chosen random arrangement, thereby generating waypoints that incrementally move the planned path out of the tightly constrained region;

[0017] Figure 7 is according to an embodiment of the present disclosure Figures 4 - 6 of a ball and slot assembly layout, where additional randomly chosen arrangements and local searches escape the tightly constrained region and find a complete path back to the initial arrangement;

[0018] Figure 8A , 8B and 8C are schematic diagrams of two components being assembled as shown in Figure 1A , 1B , 1C and 1D, where Figure 8A shows the motion planning result from the RRT technique known in the art, Figure 8B shows the motion planning result from the RRT-Connect technique known in the art, and Figure 8C shows the motion planning result from the technique of the present disclosure; and

[0019] Figure 9A flowchart of a method for robotic component assembly motion planning according to an embodiment of the present disclosure, which uses a disassembly-based process with local search to escape tightly constrained regions. DETAILED DESCRIPTION

[0020] The following discussion of embodiments of the present disclosure related to disassembly-based assembly motion planning techniques using local constraint-escape search is exemplary in nature and is in no way intended to limit the disclosed devices and techniques or their application or use.

[0021] People are good at assembling components into complete products and can use vision, touch, and intuition to assemble components together even when the assembly is tight and the required motions are complex. However, for reasons of cost-effectiveness and consistent product quality, it is desirable to use robots for such repetitive tasks and to eliminate the tedious and repetitive motions of human workers. It is known to use humans to teach robots the correct assembly sequences and motions. Unfortunately, human teaching of assembly motions is time-consuming and expensive because an accurate motion sensing environment must be established to achieve accurate motion capture of the assembly sequence. In addition, human assembly of components typically involves many trial-and-error processes, characterized by small movements to fit the components into their correct engagements. These small trial-and-error human motions are not desirable to include in the robot motion program. For these reasons, computer-based assembly motion planning techniques have been developed to replace human teaching.

[0022] Figure 1A , 1B , FIGS. 1C and 1D are schematic diagrams of two components being assembled, where traditional robotic motion planning techniques are very inefficient and highlight the need for improved motion planning techniques. Many products are designed with components having integral features that are used to ensure proper assembly and alignment with mating components. These features - such as pins and holes, protrusions and slots, nested geometries, etc. - require the assembly of components to follow a very precise motion sequence. This type of situation is shown in Figure 1A , 1B , FIGS. 1C and 1D.

[0023] A first component 110 is to be assembled onto a second component 120. Components 110 and 120 can be any two components, such as two parts of a computer chassis as shown here. During assembly, component 120 is fixed in place, and component 110 moves along a path to be assembled into its final position where it mates with component 120. Finding the motion path of component 110 relative to the fixed position of component 120 is the goal of the motion planning calculation. The geometries of components 110 and 120 are known and are typically provided as inputs to the motion planning calculation in the form of CAD solids or surface models.

[0024] Figure 1A Shows components 110 and 120 in their initial positions before assembly. Figure 1D Shows components 110 and 120 in their final positions after assembly. Figure 1B and 1C Shows different views of components 110 and 120 in intermediate positions during assembly, where the movement of component 110 along path 130 with a specific approach angle is evident. Figure 1A Shows a three - dimensional spatial grid 140, which shows the geometry and arrangement of the components during the calculation of the movement in a common fixed coordinate system. Throughout the following discussion, the "arrangement" of a component should be understood to refer to the position and orientation of the component in a defined coordinate system. That is, the arrangement of component 110 relative to the fixed component 120 includes all six degrees of freedom (DOF) - three positions and three rotation angles, where the rotation angles can be defined as yaw / pitch / roll about a local coordinate system, or Euler angles or any other suitable angular orientation convention.

[0025] Components 110 and 120 include several integral features that ensure the components fit together tightly and are properly aligned. The protrusion 112 at one end of component 110 must fit into the slot 122 on component 120. Similarly, the protrusion 124 on component 120 must fit into the slot 114 on component 110. At Figure 1C the other end of component 110 visible in, the protrusion 116 must fit into the slot 126 in component 120.

[0026] It is evident from these schematic diagrams that assembling component 110 onto component 120 must follow an exact motion sequence, including tilting component 110 to a specific angle, moving component 110 such that at least one of the respective protrusions 112 starts to engage with one of the respective slots 122, and tilting it back to a horizontal orientation while sliding component 110 slightly forward. For a robot to perform the assembly operation, this type of exact six - degree - of - freedom motion planning poses a significant challenge to known motion planning techniques.

[0027] The prior art of computer-based assembly motion planning has been disclosed, which is based on the Rapidly-exploring Random Tree (RRT) method. The RRT quickly and randomly builds a space-filling tree of nodes until it finds a path from the initial configuration to the final configuration. Given the mathematical models (surface or solid models) of two parts to be assembled and their initial and final configurations relative to each other, the RRT can be used to find the assembly motion sequence. Each node in the RRT motion planning must be evaluated and determined to be collision-free in order to be included as a valid or feasible waypoint. However, due to the tightly constrained target configurations of the two parts, the chance for the RRT to find a path waypoint that defines a motion sequence with exactly the correct relative position and orientation of the parts is very low. Therefore, it has been found that RRT-based techniques for assembly planning of tightly constrained parts are ineffective.

[0028] Other known computer-based assembly planning systems are rule-based and use geometric constraints and rules, such as planar and cylindrical constraints on the relative component configurations. This type of system can be more effective than RRT-based techniques, but typically can only handle simple component geometries due to the need to define rules for each potential interface.

[0029] To more intuitively describe the challenges of the known RRT-based motion planning techniques and the advantages of the techniques of the present disclosure, the following several figures will be used to illustrate some key concepts of motion planning and two-dimensional constraint compliance.

[0030] Figure 2 is a two-dimensional schematic diagram of a simple assembly of a ball into a slot, showing the loosely constrained nature of the initial configuration and the tightly constrained nature of the target configuration, as commonly encountered in part assembly operations. The ball 210 is shown at the initial configuration 210A at the upper left and at the target configuration 210B at the bottom center. In the target configuration 210B, the ball 210 is seated in the slot 220 of the structure 230. The ball 210 fits into the slot 220 with a very small clearance.

[0031] The workspace 240 (the entire grid) is defined as the configuration space in which the ball 210 can move relative to the structure 230. Near the initial configuration 210A and elsewhere above the slot 220, the workspace 240 is characterized by a loosely constrained region 250 where the ball 210 can move freely without interfering with the structure 230. In contrast, near the target configuration 210B and elsewhere near the slot 220, the workspace 240 is characterized by a tightly constrained region 260 where the ball 210 must be almost perfectly centered in the slot 220 to avoid interfering with the structure 230. Figure 2 The imaginary (dotted) image of the ball 210 in shows the only feasible positions of the ball 210 within the tightly constrained region 260.

[0032] Figure 3 is according to techniques known in the art Figure 2 schematic illustration of a ball and slot assembly layout, where it is difficult for a random tree motion planning method to find a path into a tightly constrained region. Figure 3 shows the same workspace 240 as in Figure 2 Including a horizontal axis 310 and a vertical axis 320 to show that the motion of the workspace 240 and the ball 210 is geometrically mapped in two dimensions in this case.

[0033] Starting from the initial arrangement 210A, a conventional RRT motion planning routine attempts to find a path to the target arrangement 210B. As understood by those skilled in the art, the RRT selects random waypoints in the arrangement space and evaluates each point for feasibility (i.e., avoiding collisions with the structure 230). Feasible waypoints are saved as nodes in the tree 330, and branches grow from each node until an infeasible waypoint is evaluated, which terminates that branch.

[0034] In Figure 3 it can be seen that the tree 330 grows to completely fill the loosely constrained region 250. However, after multiple iterations, the RRT technique fails to find a path from the initial arrangement 210A to the target arrangement 210B. To find a path to the target arrangement 210B, the nodes on the tree 330 would have to precisely land on the slot centerline 340 in the loosely constrained region 250, and then the waypoints would have to attempt to precisely land on the slot centerline 340 in the tightly constrained region 260. Even if the tree 330 grows quickly, the chance that any single branch finds its way into the slot 220 is so low that the RRT motion planning technique is extremely slow.

[0035] In view of the inefficiency of the above RRT technique, there is a need for an improved robotic motion planning technique for component assembly operations that can handle the complex geometries typical of actual components and can also effectively plan the assembly motion for tightly constrained components.

[0036] Figure 4 is according to an embodiment of the present disclosure Figures 2 - 3 schematic illustration of a ball and slot assembly layout, where motion planning starts from the target arrangement 210B, randomly selected waypoint arrangements are selected, and a plurality of local samples are defined. Compared with the RRT-based method shown in Figure 3 the technique of the present disclosure significantly reduces the number of poorly evaluated arrangements evaluated when constructing a sequence of waypoints. When considering that collision avoidance calculations must be performed at each potential waypoint arrangement in the motion planning sequence, the significant reduction in the number of arrangements evaluated results in a more efficient motion planning process.

[0037] Figure 4 (and subsequent Figures 5 - 7 ) the same ball and slot assembly and Figures 2 - 3 the workspace 240 of shows the motion planning technique of the present disclosure. The ball 210 is shown in both the initial arrangement 210A and the target arrangement 210B, and the slot 220 as described above in the structure 230. In the method of the present disclosure, the motion planning sequence starts with the target arrangement 210B. By starting in a tightly constrained region (in this case inside the slot 220), the disclosed method avoids constructing a tree of ineffective nodes and branches in many positions (as done in the Figure 3 shown in the prior art RRT-based).

[0038] Starting from the target arrangement 210B, the method randomly proposes waypoints 410 in the arrangement space. In Figures 4 - 7 the case, the arrangement space is the two-dimensional workspace 240, and the rotation of the ball 210 is not shown because it is a spherical ball, but in the general case, the arrangement space is three-dimensional, and each arrangement is defined in all six degrees of freedom (DOF), i.e., three translations and three rotations.

[0039] After randomly defining the waypoint 410, a local sample set 420 is evaluated in the local vicinity of the existing waypoint closest to the proposed waypoint. In this case, the proposed waypoint is the target arrangement 210B. The local sample set 420 is configurable in terms of quantity ( Figure 4 shows 16 of them) and the distance in the arrangement space from the existing waypoint (arrangement 210B). In the general 3D case, the maximum distance in the arrangement space will include the maximum translation distance (such as 10% of the size of the workspace 240) and the maximum angular rotation (such as 10°). These maximum ranges of the local samples in the arrangement space are merely exemplary and can be configured to any suitable value. Each arrangement in the set 420 is randomly defined within the bounds of the maximum range of the local arrangement space. Then, each arrangement in the local sample set 420 is evaluated to determine whether it meets the collision-free requirement. The idea is that by starting in a tightly constrained region and evaluating many local samples in the nearby range, the chance of finding one or more feasible (collision-free) arrangements and using these feasible arrangements to construct the motion plan is increased.

[0040] Figure 5 is according to an embodiment of the present disclosure Figure 4 of the ball and slot assembly layout schematic diagram, where the local sample arrangements are evaluated to identify which sample is collision-free and closest to the randomly selected arrangement. Figure 5Shows the next step in the disclosed method, which evaluates each arrangement in the local sample set 420, eliminates any arrangement in the set 420 that is not collision - free (i.e., infeasible), and then selects the feasible arrangement that is closest to the proposed waypoint 410.

[0041] A subset 510 of the individual arrangements in the set 420 is infeasible because the ball 210 interferes with the structure 230 to the left of the slot 220. Similarly, a subset 520 of the individual arrangements in the set 420 is infeasible because the ball 210 interferes with the structure 230 to the right of the slot 220. Among the remaining arrangements in the set 420, the arrangement point 530 is identified as the one closest to the proposed waypoint 410 in the arrangement space. Thus, the arrangement point 530 becomes the confirmed waypoint in the motion planning. In Figure 5 a two - dimensional space, the meaning of "closest" is simply the minimum distance in the plane. In the three - dimensional arrangement space of an actual component, a defined calculation can be used to identify the closest point, which includes the translational distance (the square root of the sum of the squares of the differences in x / y / z coordinates) and a rotational "distance" with an appropriate weighting factor (such as the square root of the sum of the squares of the differences in orientation angles).

[0042] Figure 6 is according to an embodiment of the present disclosure Figures 4 - 5 Schematic diagram of a ball - and - slot assembly layout, where additional local sample arrangements are evaluated relative to the same previously selected random arrangement, resulting in waypoints that incrementally move the planned path out of tightly constrained regions. Figure 6 Shows the next step in the disclosed method, which repeats the evaluation of the local samples as Figures 4 - 5 shown, but the local samples are now defined around the existing waypoint closest to the proposed waypoint, in this case the proposed waypoint is point 530. Thus, a new local sample set (e.g., a quantity of 16 as described previously; not shown in the figure for visual clarity) is defined around point 530, and a point is selected from this new local sample set that is feasible (collision - free) and closest to the same proposed waypoint 410 as described above. This results in a new confirmed waypoint 610.

[0043] A new local sample set (e.g., a quantity of 16 as described above) is then defined around the new confirmed waypoint 610, and a point is selected from this new local sample set that is feasible (collision-free) and closest to the same proposed waypoint 410 as described above. This results in another new confirmed waypoint 620. This process (evaluating the local sample set and selecting a feasible local sample that moves towards the proposed waypoint) can be repeated several times using the same proposed waypoint 410 before defining a new proposed waypoint. The number of times the local samples are evaluated before defining a new proposed waypoint is configurable and can have a value such as 3, 5, 10, or some other number. For visual effect, the distances of waypoints 530, 610, and 620 from the centerline of the slot 220 are exaggerated.

[0044] The idea behind evaluating the local sample set and selecting a feasible local sample that moves towards the proposed waypoint and doing this multiple times for a given proposed waypoint is to "slide" the moving component (ball 210) out of the tightly constrained region (slot 220) of the fixed component (structure 230). As Figure 6 can be seen, waypoints (points 530, 610, and 620) are accumulated along the direction of moving out of the slot 220 towards the open space. The purpose of this motion planning method is to escape from the tightly constrained region 260 and enter the loosely constrained region 250, from where a path back to the initial arrangement 210A can be easily found due to the lack of constraints.

[0045] After adding waypoint 620 to the motion plan, a new random waypoint arrangement can be proposed to replace the proposed waypoint 410. The newly proposed waypoint can be anywhere in the arrangement space of the workspace 420. Then a local sample set around waypoint 620 is defined and evaluated to determine which local sample is feasible and closest to the newly proposed waypoint.

[0046] Figure 7 is according to an embodiment of the present disclosure Figures 4 - 6Schematic of the ball and slot assembly layout, where additional randomly selected arrangements and local searches escape the tightly constrained region and find a complete path back to the initial arrangement. In the above manner, the proposed waypoints are defined, and multiple sets of local samples are evaluated, and the process is repeated several times to accumulate each confirmed feasible waypoint far from the target arrangement 210B. The process is repeated several times as needed to escape the tightly constrained region 260 (escape the slot 220), resulting in a set of confirmed waypoints 710 moving upward along the slot 220. Finally, the waypoints 720 in the loosely constrained region 250 will be confirmed. Thereafter, the proposed waypoints are defined, and local samples that can be located anywhere in the loosely constrained region 250 and can easily find their way back to the initial arrangement 210A are evaluated. This is shown in Figure 7 the upper part of, and generates additional waypoints 730, 732, 734, and 736.

[0047] In Figure 7 it can be seen that a motion plan has been defined, where the ball 210 travels upward from the target arrangement 210B along the set of points 710 in the slot to the point 720, then through the waypoints 730, 732, 734, and 736, and back to the initial arrangement 210A. To perform the desired operation, which is to move the ball 210 from the initial arrangement 210A to the target arrangement 210B, this motion plan needs to be reversed. Additionally, it is beneficial to prune unnecessary motions from the action sequence. For example, it can be quickly determined that the motion sequence can advance collision - free from the initial arrangement 210A to the waypoint 720 in the loosely constrained region 250, thus eliminating the wasted motions through the waypoints 730 and 736.

[0048] The final motion sequence 740 can be used to effectively move the ball 210 from the initial arrangement 210A to the waypoint 720, and then through some (not necessarily all) of the waypoints in the slot 220 until reaching the target arrangement 210B.

[0049] Figures 4 - 7 The foregoing discussion illustrates the motion planning technique of the present disclosure in terms of a two - dimensional ball and slot assembly and the workspace 240. Using a two - dimensional example to illustrate the disclosed technique is only for clarity and easy understanding. It should be understood that the techniques applied to Figures 4 - 7 each waypoint and arrangement shown in two dimensions (x and y positions) can be applied to the arrangement of one 3D component relative to another in three dimensions and six DOFs (x / y / z positions and three rotation angles). Similarly, Figures 4 - 7 the simple collision avoidance calculation (whether the ball 210 interferes with the structure 230) corresponds to the minimum distance and collision avoidance calculation between two 3D geometric models. These techniques provide an effective method for planning a motion sequence for assembling one component with another.

[0050] Figure 8A , 8B and 8C are schematic diagrams of the assembly of two components as shown in Figure 1A , 1B , 1C, and 1D, where Figure 8A shows the motion planning results from the RRT technology known in the art, Figure 8B shows the motion planning results from the RRT-Connect technology known in the art, and Figure 8C shows the motion planning results from the technology of the present disclosure. In head-to-head calculations, the technology of the present disclosure gives effective motion planning faster and more efficiently than the RRT or RRT-Connect technologies.

[0051] Figure 8A shows components 110 and 120 in their initial arrangements, as well as partial calculation results using the RRT method. The RRT method is the slowest because it evaluates a large number of ineffective waypoints, and each branch of the tree has a very low probability of finding a route into the tightly constrained areas of the assembly. This is shown by the various nodes and branches of the tree indicated at reference numeral 810.

[0052] Figure 8B shows components 110 and 120 in their initial arrangements, as well as partial calculation results using the RRT-Connect method. The RRT-Connect method starts its tree construction from the initial and target arrangements until they are connected. This method is more effective than the RRT, but still much slower than the method of the present disclosure. This is shown by the various nodes and branches of the tree indicated at reference numeral 820.

[0053] Figure 8C shows components 110 and 120 in a partially assembled arrangement because the motion planning 830 has been calculated by the technology of the present disclosure and is being executed by the robot, while the RRT and RRT-Connect methods are still in the process of calculation. These figures - the RRT and RRT-Connect methods take longer to find the assembly motion sequence than the method of the present disclosure - reflect the results of the above actual head-to-head calculations.

[0054] Figure 9FIG. 900 is a flow chart of a method for robotic component assembly motion planning according to an embodiment of the present disclosure, the method using a disassembly-based process with local search to escape tightly constrained regions. At block 902, geometric models of two components are provided, along with an initial configuration and a target configuration. As described above, the geometric models can be solid or surface models of components from a CAD system, and the initial and target configurations define the position and orientation of the components in a common fixed coordinate system. Typically, one of the components does not move, while the other component moves from its initial configuration to the target configuration, where it is assembled with the stationary component; this is consistent with Figures 2 - 7 the ball-in-slot example of FIG. 1 and the computer chassis assembly example of FIGS. 1 and 8.

[0055] At block 904, starting from the target configuration, random waypoints are proposed in the configuration space. The random waypoint and all other samples and waypoints represent the movement of one component (moved by the robot) relative to the other component (position-fixed) during the assembly operation. At block 906, a set of multiple local samples is defined around the existing waypoint closest to the proposed waypoint, the existing waypoint being initially the target configuration. The number of local samples and their distance from the current working point in the configuration space are configurable parameters in the computer implementation of the method.

[0056] At block 908, each of the multiple local samples is evaluated to determine its collision-freeness by positioning the geometric models of the components according to the sample configurations and checking for interference. At block 910, one of the multiple local samples is selected that is feasible (collision-free) and closest to the randomly proposed waypoint in the configuration space. At decision diamond 912, it is determined whether the waypoint or local sample point is consistent with (or within a tolerance of) the initial configuration. If no waypoint has yet been defined in the initial configuration, the process continues to decision diamond 914.

[0057] At decision diamond 914, it is determined whether a predetermined threshold number of local sample loops has been reached. If the number of local sample loops has not yet been reached, the process loops back to block 906, where a new set of local samples is defined around the existing waypoint closest to the proposed waypoint, the existing waypoint now being the local sample point selected at block 910. Then, the feasibility of the new set of local samples and their proximity to the same waypoint previously proposed at block 904 are evaluated. The predetermined threshold number of local sample loops is a configurable parameter and can be set to a number in the range of, for example, three to ten, to achieve optimal results.

[0058] If the number of local sample cycles has been reached at decision diamond 914, the process loops back to box 904 where a new random waypoint is proposed. A new set of local samples is then defined around the existing waypoint closest to the proposed waypoint. Each of the plurality of local sample points selected at box 910 becomes part of the saved waypoint path. The loop returning from decision diamond 914 to box 904 or box 906 continues until the waypoint path finds its way back to the initial configuration provided at box 902. When the initial configuration is reached at decision diamond 912, the process moves to box 916. At box 916, the motion sequence is trimmed to eliminate wasted motion and reversed to provide a final motion sequence from the initial configuration to the target configuration (i.e., the final motion sequence is in the direction of component assembly, where the planning calculations are performed in the direction of disassembly). The process ends at endpoint 918.

[0059] Trimming the motion sequence at box 916 includes taking shortcuts around any path points that unnecessarily increase the path distance (when there is no need to avoid obstacle collisions). Trimming the motion sequence can also include reducing the number of path points in regions with very dense point spacing, as explained with respect to Figure 7 The final motion sequence is then made available to the robot controller, which will grasp a component (i.e., component 110) and move it along the final motion sequence to assemble it to another component (i.e., component 120). The computational steps of flowchart 900 can be executed on any suitable computer / processor, and the final motion sequence is provided to the robot controller for execution. Alternatively, the computational steps of flowchart 900 can be executed on the robot controller itself.

[0060] It is important to remember that the final motion sequence for moving one component relative to another is not just a series of points, but a series of three positions and three orientations in the configuration space. That is, each waypoint, random sample, and saved point discussed above is a complete six-degree-of-freedom configuration (pose) of the movable component in the workspace coordinate system. The robot performing the assembly manipulates the movable component using the translations and rotations necessary to match each sequence configuration in the motion sequence.

[0061] Throughout the foregoing discussion, various computers and controllers have been described and implied. It should be understood that the software applications and modules of these computers and controllers are executed on one or more computing devices having a processor and a memory module. In particular, this includes the processors in each of the robot controllers and other computers (if used) discussed above. Specifically, the processors in the controllers and / or other computers are configured to perform the disassembly-based assembly motion planning techniques in the manner described and illustrated throughout the foregoing disclosure using local constraint - escape search.

[0062] As described above, the disclosed technique of disassembly-based assembly motion planning using local constraint-escape search provides significant advantages over prior art methods. By starting from the target configuration and using local search loops to escape tightly constrained regions, the disclosed technique is much faster and more efficient than prior art RRT-based assembly motion planning techniques, and also applicable to complex part shapes, unlike prior art rule-based assembly motion planning techniques.

[0063] Although multiple exemplary aspects and embodiments of the disassembly-based assembly motion planning technique using local constraint-escape search have been discussed above, those skilled in the art will recognize its modifications, permutations, additions, and sub-combinations. Accordingly, the appended claims and the claims hereafter introduced are intended to be construed to include all such modifications, permutations, additions, and sub-combinations in their true spirit and scope.

Claims

1. A method for planning the assembly movement of robot components, characterized in that, The method includes: providing geometric models of a first component and a second component to be assembled, an initial arrangement and a target arrangement of the first component relative to the second component; defining the target arrangement as a first stored waypoint; randomly selecting a proposed waypoint in the arrangement space; defining a set of a plurality of local sample arrangements around the stored waypoint closest to the proposed waypoint; evaluating each of the plurality of local sample arrangements to identify a local sample arrangement that is collision-free and closest to the proposed waypoint in the arrangement space, and adding the identified local sample arrangement as a stored waypoint; when adding a stored waypoint at the initial arrangement, reversing the sequence of the stored waypoints to provide the assembly motion; when the local sample loop count quota has not been reached, returning to defining the set of the plurality of local sample arrangements; and returning to randomly select a proposed waypoint arrangement.

2. The method according to claim 1, characterized in that, It further includes removing any one of the stored waypoints from the assembly motion, and the any one of the stored waypoints can be removed without causing a component interference situation during the assembly motion.

3. The method according to claim 1, characterized in that, It further includes providing the assembly motion to an industrial robot through a robot controller, and using the assembly motion by the robot to assemble the first component to the second component.

4. The method according to claim 1, characterized in that, Each waypoint and sample arrangement in the arrangement space includes a three-dimensional position of the first component relative to the second component and three orientation angles.

5. The method according to claim 1, wherein The set of the plurality of local sample arrangements is within a predetermined range of distance and rotation angle of the stored waypoint closest to the proposed waypoint.

6. The method according to claim 1, wherein Evaluating each of the plurality of local sample arrangements to identify a local sample arrangement that is collision-free includes performing a mathematical interference check on the geometric models of the first component and the second component.

7. The method according to claim 1, characterized in that Evaluating each of the plurality of local sample arrangements to identify a local sample arrangement that is closest to the proposed waypoint in the arrangement space includes using a weighted sum of a translation difference term and a rotation difference term.

8. The method according to claim 7, characterized in that, The translation difference term is the sum of the squares of the differences in the x, y, and z positions of the sample arrangement relative to the proposed waypoint, and the rotation difference term is the sum of the squares of the differences in the respective angular positions of the sample arrangement relative to the proposed waypoint.

9. The method according to claim 1, characterized in that, The set of the plurality of local sample arrangements has a size within the range of ten to thirty local sample arrangements.

10. The method according to claim 1, characterized in that, The local sample loop count quota is predetermined and has a value within the range of three to ten.

11. The method according to claim 1, wherein Adding a stored waypoint at the initial arrangement includes adding the stored waypoint within a predetermined range of distance and rotation angle of the initial arrangement.

12. The method according to claim 1, characterized in that, The assembly motion includes a sequence of waypoints starting from the initial arrangement and ending at the target arrangement, where each waypoint is a six-degree-of-freedom arrangement of the first component relative to the second component.

13. A method for planning the assembly movement of a robotic component, characterized in that, The method includes: providing geometric models of a first component and a second component in an initial arrangement and a target arrangement; defining the target arrangement as a first stored waypoint; evaluating a set of multiple local samples around the stored waypoint closest to the proposed waypoint to identify a sample that is collision-free and closest to the randomly proposed waypoint and storing it as a waypoint; repeating the evaluation of the set of multiple local samples relative to the same randomly proposed waypoint and new randomly proposed waypoints until the stored waypoint is within the tolerance range of the initial arrangement; and reversing the sequence of the stored waypoints to provide the assembly motion.

14. The method according to claim 13, wherein It further includes removing any one of the stored waypoints from the assembly motion, and any one of the stored waypoints can be removed without causing a component interference situation during the assembly motion.

15. The method according to claim 13, wherein Each waypoint and local sample includes the three-dimensional position of the first component relative to the second component and three orientation angles.

16. The method according to claim 13, wherein The multiple local samples are within a distance and rotation angle of a predetermined range of the stored waypoint closest to the proposed waypoint.

17. The method according to claim 13, wherein Evaluating a set of multiple local samples to identify a sample that is collision-free includes performing a mathematical interference check on the geometric models of the first component and the second component.

18. The method according to claim 13, characterized in that, Evaluating a set of multiple local samples to identify a sample that is closest to the randomly proposed waypoint includes using a weighted sum of a translation difference term and a rotation difference term.

19. The method according to claim 18, wherein The translation difference term is the sum of the squares of the differences in the x, y, and z positions of the sample relative to the proposed waypoint, and the rotation difference term is the sum of the squares of the differences in the respective angular positions of the sample relative to the proposed waypoint.

20. The method according to claim 13, wherein The assembly motion includes a sequence of waypoints starting from the initial arrangement and ending at the target arrangement, where each waypoint is a six-degree-of-freedom arrangement of the first component relative to the second component.

Images