Reconfigurable Swarm Fixtures - IEEE Xplore

Reconfigurable Swarm Fixtures Rezia Molfino, Matteo Zoppi, Dimiter Zlatanov PMAR Laboratory, DIMEC, University of Genoa Via all’Opera Pia 15a, 16145 Genoa, Italy [molfino, zoppi, zlatanov]@dimec.unige.it

Abstract— The paper presents the novel concept of a reconfigurable swarm fixture. In existing fixturing systems the configuration of the supports cannot be modified significantly without removing the workpiece; reconfiguration is usually manual and time consuming. In contrast, in the system we are currently developing, supports are mobile agents controlled to behave as a swarm and able to reposition automatically during the manufacturing of the workpiece. The goal is to create a simple to operate, modular, and cost-effective fixturing system with a relatively small number of supports, allowing the uninterrupted machining of large thin parts like those used in the production of aircraft and automobiles.

1. INTRODUCTION Components made of thin metallic or composite sheets are ubiquitous in automotive and aircraft manufacturing and are becoming increasingly common in other sectors. This is in keeping with current trends favouring life-cycle design and sustainable production. Economy of materials and higher aesthetic standards cause a reduction in the thickness of the sheets used and an increase in their geometrical complexity. (Moreover, sometimes non-standard shapes are adopted to achieve stiffness equivalent to that of thicker material.) These requirements affect changes in the manufacturing process, e.g.: accurate grooving and windowing operations become more common; continuous welding replaces spot welding; and round holes at precise locations are preferred to the slotted holes traditionally used for relative adjustment in assemblies. Demand for higher product quality leads to higher labour costs and an increased desirability of automation. On the other hand, short time-to-market, production in small and variable batches, as well as the mass customization for some products, all call for increased flexibility. Thus, to keep in step with changing requirements, manufacturing equipment must meet the combined challenge for fuller automation and higher flexibility. Usually thin sheets are formed (by pressing or other forming technology) and then manufactured (deburred, milled, holed, contoured, welded) [1,2]. Flexible equipment for accurate machining and welding is available. The critical operation is fixturing: thin compliant walls require the contact and pull forces between fixture and workpiece to be suitably distributed on large areas; complex surface geometries call for com-

730

plex mating fixture shapes which are difficult to realize with standard flexible fixtures. 2. IN USE FIXTURING SYSTEMS Here we briefly review existing flexible fixturing systems [3], and evaluate their merits with regard to the flexible automated manufacturing of thin-walled parts. 2.1. Flexible fixture systems (FFS) 2.1.1. Modular flexible fixture systems (MFFS) MFFSs are quite popular in industry. Their basic elements, a base plate, locators, clamps, and connections, can be combined in various configurations [4-6]. These systems can be adapted to various workpieces but their initial cost is often high while configuration is complex and time consuming. MFFSs can be further classified on the basis of their adjusting mechanism: (a) with manual reconfiguration (sometimes software assisted); (b) adjusted by one or more separate devices/robots; (c) with embedded actuators (in each locator/clamp). All such fixtures still require some human intervention to reconfigure. 2.1.2. Single structure flexible fixture systems (SSFFS) SSFFSs include adaptive and phase-change FFSs. The former adjust to the workpiece by means of internal mechanisms [7]; the latter use a phase-change (e.g., electro/magnetorheological) fluid to reproduce the shape of the workpiece. A relevant development is the system with pin-bed layout described in [8]: movable pistons fixed to a base and immersed in the phase-change fluid are operated to adapt the fixture surface to the desired workpiece shape; the fluid phase is then changed to solid to obtain a uniformly stiff support. Phasechange fluid surrounding the workpiece is also used in the conformable clamp [9] developed at MIT. In recent years SSFFSs with matrices of supports have become available on the market. These “bed-of-nails” systems, aimed primarily at aeronautics production, represent a significant advance compared to conventional fixtures. Modig developed the Universal Holding Fixture for Skins It is used for clamping 3D formed sheets during milling, holing, and laser scribing. The system uses a grid of posts, adjusted, row by row and off-line, by a gantry machine working either from below or from above. At the end of each post is a

J.S. Dai, M. Zoppi and X. Kong (eds), ASME/IFToMM International Conference on Reconfigurable Mechanisms and Robots

Reconfigurable Swarm Fixtures

ball with a vacuum cup to support and hold the workpiece. Posts can be active during the machining process allowing retraction under the cutting tool. Kostyrka GmbH offers a hydraulically actuated support, which can be used to assemble a multi-post system for aircraft and aerospace applications. Yet another example of this type is CompoFlex by Jobs. It is distinguished by the use of a cradle-shaped base bench composed of three hinged plates. The system is automatically controlled off-line and the set up time is in the order of 20 minutes. Head suction cups are used to lock the workpiece. 2.1.3. Robotic fixtureless assemblies (RFAs) RFAs replace traditional fixtures by sensor-guided robots equipped with programmable grippers that can cooperatively hold the workpiece [2,10]. Advantages are that different parts can be produced with one work cell and changeovers to other workpieces can be done quickly. Drawbacks are complexity, limited number of robots (and thus holding grasps), high dependence on software. Computer-aided fixture design allows the automatic generation of fixture configurations for both MFFS and SSFFS. Computer-aided fixture design verification (CAFDV) [11-13] improves MFFS and SSFFS by increasing the level of flexibility but still the adaptation process, especially for MFFSs, needs time and is hard to automate. Moreover, these features are available only for sufficiently rigid workpieces [14].

Fig.1 – Conceptual schematic of self reconfigurable swarm fixture: (1) part; (2) manufacturing equipment; (3) bench; (4) agent; (5) support head; (6) positioning mechanism; (7) mobile bases; (8) concentration of agents in the manufacturing region

3. SELF RECONFIGURABLE SWARM FIXTURES No fixture type in the state of the art has simultaneously the advantages of short reconfiguration time, easy set-up, adaptability to large shape ranges, and low cost compared to performance. All have to be reconfigured off-line, before loading the workpiece. An emerging solution is to use a limited number of mobile supports that can self-reposition. It is required that repositioning is fast enough to match the timing of manufacturing operations and that the mobile supports can re-establish their con2.2. In use fixtures for skins tact with the supported part without disturbing the overall part Currently, the fixtures for thin-walled parts with complexconfiguration (by causing local displacements or vibrations). surface geometries are either: (i) mould-like (part specific and The control of the repositioning of the mobile supports can not reconfigurable); (ii) partially reconfigurable with limited be efficient and robust if they are implemented as a swarm: number of support elements that can be manually relocated; each support is an agent following simple rules; the desired (iii) self reconfigurable with matrices of support elements with task is carried out by coordinated action. A system based on variable configuration. this new concept is referred to as a self reconfigurable swarm The fixtures traditionally used in aeronautic manufacturing fixture, Fig.1 [15]. are large moulds reproducing the shape of the skin to be supIn the course of its production, a workpiece often underported. The mould surface is equipped with vacuum suction goes a sequence of manufacturing operations, all at the same chambers and channels for holding the skin. machining centre, but each with specific (and possibly timeMFFSs with a limited number of supports, to be reposivariable) fixturing requirements. During this process, the motioned manually or by separate robots, have been proposed [6], bile supports continuously concentrate in the region where but are not widely used. Since fixturing requirements vary manufacturing is currently performed ((8) in Fig.1) without during the different machining operations required on a single displacing the part from the fixture. Thus the zone near the part, it becomes necessary to reposition the supports, intercurrent position of the machine tool is actively supported, rupting the production process. while regions further away are just held in place by a miniOne way to avoid this problem is to use an SSFFS of the mum number of agents. pin-bed type, with a matrix of support elements, which proThe idea of a swarm fixture develops the RFA concept at vides support comparable to a mould-like fixture. However, an even higher level of modularity and flexibility. Moreover, such systems are yet to gain wide acceptance. Their main disthe proposed system merges these advantages of RFAs with advantages are high cost, and a lack of modularity, which those of MFFSs, namely: ability to distribute the support acmakes them difficult or inefficient to use for parts of differing tion; adaptability to workpiece shapes in a larger range; and sizes. These problems stem from the higher complexity due higher stiffness of the provided support. to the need for a very large number of supports, each of them Benefits are particularly evident for large workpieces with actuated to allow automatic reconfiguration between work3D surface geometry, like those used to manufacture airplane pieces. (If supports are unactuated, separate robotic manipulafuselage or automobile bodies. In particular, the usual large tors, and more time, are needed to perform the reconfiguramould-like fixtures in aircraft production are very costly to tion). manufacture and store while their flexibility is null (each

731

Rezia Molfino, Matteo Zoppi and Dimiter Zlatanov

mould is used for only one side of one workpiece). Available SSFFSs with post-matrix architectures, while highly flexible, are heavy, complex and even more costly. At every stage in the manufacturing process, only a small portion of the worked surface needs the mould-like highrigidity support of in-use systems. In the proposed solution this inherent redundancy is eliminated, while still ensuring that the machining is performed in one set-up. The use of swarm intelligence yields a solution that is intrinsically agile and economical.

Fig.2 – Simple swarm fixture with planar bench and one basic agent type

4. SWARM FIXTURE ARCHITECTURE The swarm fixture comprises a bench and a variable number of support agents. The bench is likely to be planar, possibly consisting of two or more articulated panels that can be arranged as a “cradle” partially surrounding the workpiece. Though it is desirable, for simplicity, all agents to have uniform design, it may be necessary to have several different types. The simplest setup, with a single solid planar bench and identical agents, is illustrated by Fig.2. 4.1. Support agents The support agent is an intelligent unit with some autonomy able to collaborate and co-operate with other similar units. It can move on the bench, anchor to it, accurately position and adapt its locator(s) (support head(s)) to the supported part. A basic agent architecture, (4) in Fig.2, may comprise one mobile base (7), one manipulator connected to the mobile base (6) and carrying one support head (5) which adapts to the local geometry of the workpiece (1) passively or actively. The architecture of the head positioning manipulator can be serial, (12) in Fig.3, as is common with MFFSs locators, or it can be parallel or complex for higher stiffness. More complex agent architectures are possible as well, Fig.3, with multiple bases, one or more heads, and different positioning mechanism architectures. A critical characteristic of the fixturing system is the size of an individual agent. The optimal choice depends on the size and stiffness of the supported part, the shape of the machined surface, as well as the types of the performed manufacturing processes. To maximize the advantage of using a swarm fixture, it is desirable to have agents that are small in relation to

732

the manufactured part. However, the need for the autonomous agents to be equipped with actuators and electronics imposes technical limitations to miniaturization. Moreover, if the manufacturing forces and torques are high the agents must be sufficiently robust (and numerous) to keep the workpiece fixed. Furthermore, workparts with variable geometry or high curvature require the agents to have large workspaces and hence large dimensions. On the other hand, if the piece is with low stiffness, the load must be distributed among multiple agents in close proximity to each other, and this is easier to achieve if the agents are small. These considerations imply that swarm fixtures should be best suited to the manufacturing of large parts subject to comparatively small forces and torques. Therefore, the aeronautic and automotive industries are ideal target applications for the proposed system. The determination of the agent architecture involves several complex decisions: the selection of the principle of locomotion (movement on the bench); the choice of an anchoring mechanism; the synthesis of the agent manipulator including its degree of freedom and actuation; the design of the agent head; a solution to the power-supply and communication problems, in particular the selection of onboard electronics components as well as the agent-bench interface. Unlike traditional static fixtures, the swarm system concept implies a dynamic setup with the number and locations of the supports changing during the execution of the manufacturing operations on the workpiece. Fixture configurations for specific CAD sheet geometries and manufacturing cycles can be planned, programmed, and verified using an offline simulator. The generated map of support locations is used by the supervisory controller. An internal simulator determines the minimum number of supports and clamps assuring correct overall position and shape of the sheet. It also derives and manages in an optimal way the concentration of support agents in the machined regions. The agent’s control system is in charge of all low-level activities as well as two higher level functions: locomotion and head repositioning. The latter involves detaching the head from the part, and its retraction at a safe distance, as well as the approach to the workpiece followed by adhesion and locking in the new location. A particularly critical operation is the head’s approach to the workpiece surface. The motion will be fast before contact is made and then slower. Since the expected distance between head and part will be known only with limited accuracy, the head has to sense the contact with the workpiece. The exact distance between the reference point and the surface remains difficult to determine after the compliant head touches the surface, because it depends on the unactuated (and generally unknown) orientation of the agent’s head. Therefore, a robust control system for the head positioning mechanism is required and a proximity sensor embedded in the head may be necessary. For controlling locomotion, the agent’s guiding system needs information on the current position and on the presence


of other agents interpreted as obstacles to avoid. Reference marks, like holes on the bench, can be used to generate stepped displacements. (Head adaptability and possibly redundant positioning make stepped agent displacements fully acceptable.) A low friction pad can be adopted on the agent base to facilitate locomotion.

9

10

11

12

13

14

Fig.3 – Alternative agent architectures with single and multiple base and support heads

4.2. Agent heads Heads adapt to the local geometry of the workpiece to support it at every repositioning. Adaptation is at two levels: head rotation, to match the (approximate) orientation of the part surface normal; and head surface deformation, to match the local part surface geometry. Head rotation can be passive, provided that the head can be locked once it has been oriented. Two rotational freedoms are sufficient, but a spherical passive joint can be adopted as well, if the head has an axis of symmetry or head torsion is not relevant. The adaptation to the workpiece’s local surface geometry is achieved by elastic compliance of the head. Head and part are forced to adhere by the adhesion system in the head. At least one portion of the head shall be stiff to provide support of the part against translation during manufacturing. This stiff portion must be sufficiently small not to inhibit adaptability and can be assumed to define a reference contact point. The rest of the head may be passively compliant, and after the workpiece is in position the deformed head provides a reaction torque constraining the rotations about the stiff contact (as well as some resultant force through the reference point). Alternatively, a head whose rigidity can be actively controlled may be used to minimize the deformation and preserve the equilibrium of the part during the process of adaptation. For this purpose, head compliance is initially increased and later, when adaptation is complete, is lowered again to provide stiff support with almost uniform contact pressure along the head-part interface [16]. Moreover, in this case the adhesion force provided by the head adhesion system is fully used to counter external (manufacturing) forces. In contrast, if the head were passively compliant, part of the adhesion force must balance the repulsive elastic force. Furthermore, the head, once stiffened, provides better constraint, both by distributing the normal reaction force over the entire head/part contact surface, and allowing higher torque to prevent the workpiece from rotating about the reference contact point. The baroplastic behaviour can be obtained using a phasechange fluid or a shape memory material [17], suitably enveloped. Alternatively, the compliant head may be designed as a sealed chamber containing particles which can be stacked

solid by applying vacuum [18-20]. The selected material must be sufficiently compliant in its soft state to deform at contact with the skin, and the phase change must occur very rapidly. The head-to-part adhesion system may use different physical principles. Adhesion and adaptation functionalities coexist in the head and the major difficulty is to accommodate both in the small overall head volume. Vacuum may be a good option as it can also be used with materials that are porous or with irregular surfaces, if sufficient air flow is provided by the vacuum generator [21,22]. Head compliance guarantees accurate adhesion of the head to the part limiting vacuum losses. Heads specially intended for ferromagnetic workpieces could use magnets. Soft materials like foam shaped elements could be blocked using retractable needles. Glue and other inuse physical principles are feasible as well according to specific workpiece characteristics. An important requirement of the adhesion system is to have negligible springback; otherwise fixturing accuracy would be reduced unpredictably. 4.3. Bench The bench provides a base surface on which the agents move and to which they can lock once they are in position. It is expected to incorporate power-supply and communication systems as well as provide the means to measure the accurate absolute positions of the agents. The design of the power and control electronics will be critical. The size of agents is limited by functional constraints coming from the task. There may be insufficient space to package control electronics and actuation onboard the agent. Solutions with part of the hardware in the bench may be adopted, but this increases the overall complexity of the swarm fixture. Holes in the bench can be also used for docking the agent when it is supporting the workpiece. A ferromagnetic bench with magnets in the agent can be used to ensure adhesion during displacement (when locking is off) or with a nonhorizontal bench surface. The geometry of the bench (planar, cradle, articulated) depends on the geometrical variability of the workpieces. The closer the bench envelops the workpiece surface, the smaller the required adaptation range of the agents. (The needed support stiffness is easier to obtain if the agents can be smaller and simpler.) On the other hand, the geometry of the bench surface affects the locomotion of the agents: adaptation to changes of bench curvature may increase complexity; articulated patchwork benches with internal edges are problematic because they make it difficult for agents to cross from one patch to another. 4.4. Modularity and scalability As Figure 4 illustrates, the swarm fixture is conceived to be intrinsically modular and scalable. Given the geometry of a family of workpieces, one selects the size and geometry of the bench, as well as the number and types of the agents. Individual agents can have different heads, varying in size, adaptability, stiffness, and adhesion principle. Ideally, the base surface (on which the agents move and dock) should be a generic bench with regularly spaced holes

733

Rezia Molfino, Matteo Zoppi and Dimiter Zlatanov

or grooves. This would require having all (electronic and possibly pneumatic) equipment onboard the agents, which may be difficult. A compromise solution can be a partially specialized bench derived from in-use fixtures. Specialization in agent types is possible but it requires additional planning rules, because agents are no longer interchangeable. For example a rule may constrain agents of different heights to different regions of the bench, avoiding interference with a curved part surface. An easy to use programming interface will be important, because it must ensure simple operation despite the intrinsic complexity of swarm systems.

piece). This machining cycle is the target application for the new fixturing system. A more ambitious further goal would be to replace the chemical etching with a third type of machining process, micro-milling. Unlike the drilling and contouring operations which are performed on one side of the sheet, the introduction of micro-milling will require the part to be machined on both sides.

Fig.5 – The parts of aircraft body selected

Moreover, micro-milling, when compared to the current machining processes, imposes different requirements on the accuracy of the fixturing system. The same five-axis CNC 5. INDUSTRIAL APPLICATION AND REQUIREMENTS machine can be used for all processes, and during each the tool should be oriented along the surface normal. While drill5.1. Prototypal application ing and milling are sensitive to inaccuracies in the tangential The new swarm fixture is being developed with special atdirections (orthogonal to the tool), they require little precision tention to the needs of the automotive and aerospace indusalong the tool axis. For the micro-milling of pockets, the retries. For example, in the production of an aircraft a number of quirements are essentially reversed, since accuracy in the large parts, made from formed aluminium sheets, are subject normal direction needs to be very high. to processes of holing, milling, deburring and contouring. Various tools are currently used in the work cycles and for Typically, a five-axis machine tool with a single spindle is each machining forces and spindle torque are known. The used. Prior to that, chemical etching is adopted in some cases design requirements are defined by the worst-case existing to modify the thickness of some regions of the panels. conditions, i.e., the drilling and milling processes with the The production of such parts has been chosen as the protomaximum manufacturing forces/torques. Micro-milling is not typal application, around which the development, simulation currently applied but the expected forces and torque can be and testing of the system is organized. derived from known analogous processes. In particular, we have selected five representative thinDrilling applies an almost static force to the workpiece and sheet parts used in the production of the aircraft model P180 a variable torque with a peak in the end of the operation when of Piaggio Aero Industries, Fig.5. In addition to performing all the drill bit crosses. operations in the current manufacturing cycle, it is desirable to Milling applies a dynamic force and torque due to the replace the chemical etching of the panels with mechanical rhythmic impact of the tool cutters against the workpiece. Vimicro milling for cost and environmental benefits. brations may propagate along the workpiece. These vibrations The five parts have thicknesses between 1 and 2 mm and may cause the workpiece to detach from the fixture in the retheir overall sizes vary from 600x600 to 1200x3000. The gion close to the tool. Mould-type fixtures damp workpiece chemically etched pockets have depths of 0.5 to 1 mm. They vibrations more effectively than a set of supports distributed are distributed on both sides of the panels, the larger pockets below the workpiece, as in the swarm fixture and in MFFSs. being predominantly on the internal surfaces. Locking agent head rotations when the agent is supporting the 5.2. Current production cycles and fixturing requirements workpiece contributes to damp these vibrations. Micro-milling applies smaller forces and torques. However, Fixturing requirements for the five parts are now being dethe frequency of the dynamic load is higher. rived by studying the existing manufacturing cycles. Currently, after the metal sheets are formed and chemically etched they 5.3. Feasibility and specifications of the new system are machined on a solid mould-like fixture. Two types of maA typical set-up for the use of the new fixture system on the chining processes are used: drilling (of various small holes) selected parts would be as follows. The piece will be held in and milling (when making larger holes and contouring the position by a subset of locators that will remain largely static Fig.4 – Intrinsic modularity and scalability of the swarm fixture concept

734


during the cycle. They are placed in areas where no machining takes place, preferably mostly on the periphery of the part. The static supports ensure rigid body equilibrium and serve as reference posts to locate the workpiece with respect to the machine tool. The remaining agents will be highly mobile and change locations to provide additional support in areas affected by the machining process under way at each point in time. Since the curvature of the workparts is not large, they can be fixtured so that the tangent plane is close to horizontal on most of the surface. Gravity deforms the workpiece and the regions far from the reference posts tend to displace vertically. The mobile agents below the part recover the gravity deformation pushing up the surface to its ideal geometry. The need for an accurately positioned stiff portion in the agent head is now clear: the actual coordinates of the reference point (in the agent head) need to be known with accuracy and to deviate little from the programmed position, which is calculated using the ideal geometry of the workpiece. The stiffness of the agent can be different in the different directions. The highest stiffness is needed in direction normal to the workpiece surface, because during the part deformation tangential displacements are small. A parallel mechanism can be used as a locator agent to position the head and the selection of the architecture and its design of the mechanism may be remarkably influenced by these stiffness requirements. Assuming that the design of an individual agent provides sufficient accuracy and stiffness in positioning its reference point, the question arises how to position the mobile locators so that the fixture configuration provides sufficient support. Various candidate sets of stiff reference points will be generated on the basis of different hypotheses of the elastic behaviour of the part under the machining loads. Then, FEM simulations will be used to determine feasible support configurations. In turn, the analysis of these acceptable sets of locations will be used to derive more precise tolerances for the deviations of the reference points from their ideal positions. 6. CONCLUSION New approaches in multi-robot control, based on swarm intelligence, as well as the availability of intelligent materials, allow the development of novel fixtures whose support of the workpiece dynamically adapts to the changing manufacturing requirements. Support units move on a bench as a swarm of agents controlled by a supervisor. Thus, a limited number of agents can provide support along all the surface of a large workpiece. The use of sophisticated materials and control techniques yields potential savings in setup time and cost in comparison to multi-post and other existing flexible systems. ACKNOWLEDGMENT The research is developed within the SwarmItFIX project funded under the Seventh Framework Programme (Collaborative Project 214678). We acknowledge the assistance of the European Commission and the other partners of the project: Warsaw University of Technology (Poland), Piaggio Aero

Industries (Italy), Exechon (Sweden), ZTS VVÙ Košice a.s. (Slovakia), and Centro Ricerche FIAT (Italy). REFERENCES [1] Zoppi, M., Bruzzone, L., Molfino, R: “A parallel hybrid machine for automated manufacturing of body panels of the aircraft Piaggio Aero P180”, Parallel Kinematic Machines in Research and Practice. 5th Parallel Kinematic Seminar PKS 2006. (Neugebauer R. ed.). IWU, Chemnitz 2006, April 25-26, Vol. 33, pp 721-731. [2] Backes1 F., Franke1 V., Geiger1 M.: “Concurrent manufacturing of parts and tools for the sheet-metal industry”, Journal of Intelligent Manufacturing , Springer Netherlands, 9(4), 1998. [3] Bi, Z.M., Zhang, W.J.: “Flexible Fixture Design and Automation: Review, issues and future directions”, International Journal of Production Research, 39(13), pp. 2867-2894, 2001. [4] Asada, H., Andre, B.B.: “Kinematic analysis of workpiece fixture for flexible assembly with automatically reconfigurable fixtures”, IEEE Journal of Robotics and Automation, 1(2), pp. 86-93, 1985. [5] Shirinzadeh, B., Tie, Y.: “Experimental investigation of the performance of a reconfigurable fixture system”, International Journal of Advanced Manufacturing Technology, 10(5), pp. 330-341, 1995. [6] Sela, M.N., Gaudry, O., Dombre, E., Benhabib, B., “A reconfigurable modular fixturing system for thin-walled flexible objects”, International Journal of Advanced Manufacturing Technology, 13(9), pp. 611617, 1997. [7] Wallak, A.S., Canny, J.F., “Planning for modular and hybrid fixtures; Proceedings of the IEEE Int. Conference on Robotics and Automation, San Diego”, CA, 8-13 May, pp. 520-527, 1994. [8] Grippo, P.M., Thompson, B.S., Gandhi, M.V., “A review of flexible fixture systems for computer-integrated manufacturing”, International Journal of Computer-Integrated Manufacturing, 1(2), 124-135, 1988. [9] Plut, W. J., Bone, G. M., “Grasping of 3-D sheet metal parts for robotic fixtureless assembly”, Proceedings of the CSME Forum, 13th Symposium on Engineering Applications of Mechanics, Ontario, Canada, pp. 221-228, 1996. [10] Kang, Y., Rong, Y., Yang, J., Ma, W., “Computer-aided fixture design verification”, Assembly Automation, 22(4), pp. 350-359, 2002. [11] Kang, Y., Rong, Y., Yang, J., “Computer-aided fixture design verification. Part 2. Tolerance Analysis”, International Journal of Advanced Manufacturing Technology, 21(10-11), pp. 836-841, 2003. [12] Kang, Y., Rong, Y., Yang, J., “Computer-aided fixture design verification. Part 2. Tolerance Analysis”, International Journal of Advanced Manufacturing Technology, 21(10-11), pp. 842-849, 2003. [13] Zhang, W.J., Xie, S.Q., “Agent technology for collaborative process planning: a review”, International Journal of Advanced Manufacturing Technology, 32(3-4), pp. 315-325, 2007. [14] R.C. Michelini, G.M. Acaccia, M. Callegari, R.M. Molfino, R.P. Razzoli, “Dynamics of a cooperating robotic fixture for supporting automatic deburring tasks,” International Conference on Informatics and Control, St. Petersburg, Russia, June 9-13, 1997. pp. 1244-1254. [15] Patent IT2008GE00003, “Dispositivo riconfigurabile multi-agente per il sostegno di corpi a geometria complessa”, January 2008. Applicants: R. Molfino, M. Zoppi. [16] Patent WO/2006/078742, “Reconfigurable fixture device and method of use”. [17] M. Sreekumar, T. Nagarajan, M. Singaperumal, M. Zoppi, R. Molfino, “Critical review of current trends in shape memory alloy actuators for intelligent robots”, Industrial Robot, 34(4), 2007 Page: 285 – 294. [18] Patent WO2008078012, “Method and device for using a molding or setting system comprising a sealed chamber containing particles”. [19] Patent FR2870319, “Multipurpose spacer support for maintaining and positioning complexly shaped parts”. [20] Patent US7267542, “Molding apparatus and method”. [21] R.M. Molfino and M. Zoppi, “Mass-customized shoe production: A highly reconfigurable robotic device for handling limp material”. IEEE Robotics and Automation Magazine, 12(2):66–76, 2005. [22] E. Carca, M. Zoppi, and R. Molfino, “A cooperative gripper for handling and hanging limp parts”. In Advances in Mobile Robotics, 11th CLAWAR International Conference, pages 843–850, World Scientific, Coimbra, Portugal, Sept. 8-10 2008.

735