Sharing Resources over the Internet Using Network Optimisation Techniques

This paper is a contribution to a challenging problem of the global design and standardisation of industrial robotic systems using new possibilities provided by modern information technologies and Internet. The main idea is to create an international network (consortium) of different national teams of robot designers in academia and industry in Europe, Japan and Israel, each team being responsible for its own domain of research. The entire design process is decomposed into sub-processes and co-ordinated by the general aim: to achieve a scientifically motivated answer to the “Six R’s” question: “How to design the right robots, in the right quantity and the right quality, at the right place, at the right time and at the right costs”. The Internet technology provides rapid access to the up-to-date information to all users of it, inside the manufacturing company or outside and to rapidly design and manufacture different elements of the same robot in different places. The crucial problem is to share the available bounded resources over the Internet. Considering the vast possibilities offered today by new information technologies, it is surprising that most companies involved in designing and manufacturing of industrial robots still act as a collection of isolated or weakly-connected local designers and suppliers. As such, they struggle with dynamically changing global markets of robotics products designed elsewhere and are often forced to apply insufficient and outdated know-how, technologies and information. Much of the difficulties in applying new manufacturing technologies and new theoretical design concepts comes from the poor connection between university research teams in different countries, on the one hand, and between academic research and real robot manufacturing plant in practice, on the other. The objectives of the global web-based CAD of the robotic systems (GLORO_CAD) are: • Development of a global www-based intelligent decision-support system (DSS) for a joint, simultaneous and secure design of high-precision industrial robots, operating between co-operating teams and companies of the consortium located in different places. • Decomposition of the robotic cell design process into separate stages (subprocesses) which are performed by different teams and standardization of optimal technological decisions. • Development of theoretical methods of parallel groupware design permitting to use several different design processes in different places to achieve a final robotic product with the required quality. • Further development and refinement of general concepts of the FMS design (Heragu 1992, Kusiak 1987, Stecke 1983) taking into account specific features of the robotic technology. • Development of new efficient computer algorithms for the effective on-line programming, control and scheduling of industrial robots in robotic systems. At present, the most preferred application of robotic CAD systems in industry are spot and arc welding where the majority of the existing robots are employed. Along with this, the relevant software tools have been also developed for the design of robots for the paint spraying, laser beam and water jet cutting, drilling, etc.. Recently, new robotic applications have appeared such as remote telemanipulation in telemedicine, nuclear decommissioning, deep sea and space investigation. However, the basic design stages are similar for many industrial applications and independent of the particular technology. We will distinguish the following basic stages: • Creating 3D model of the robotic workcell which includes: importing the workpiece geometry from other CAD systems or creating a new 3D model by using robotic CAD integrated tools and built-in libraries of robots, fixtures and related equipment; technological tool selection or design based on geometry and process attributes, optimal robot placement to avoid collisions and minimise the cycle time. • Calibrating the simulation model which include: calibration of the tool centre point, possible servo-axes and other auxiliary axes; determine precise position of all workcell components; calibration of robot internal geometric parameters, if needed; • Creating the off-line programs in a neutral or “robot-native” language including generation of the tool paths from the workpiece graphic data, automatic path segment sequencing; collision-free path optimisation and generation of the program listing. • Verification of the created programs: reachability verification; collision and near-miss detection, checking the joint limits, speeds and accelerations; cycle time analysis; checking of the tool orientation; cable assembly simulation to determine rubbing points and twists. • Downloading the created programs to the robot controller. Within the networked designing/manufacturing paradigm, the designing and manufacturing organisations are viewed as coalitions of responsible units (networked industries, supply chains, firms, plants, cells, design and research groups, individuals and equipment). Robotic systems in modern manufacturing represent sophisticated networked equipment (robots-manipulators, workstations, workpieces, positioners, conveyors, etc.) connected by various material and information flows. Resource allocation and optimisation problems arising in such robotic systems are often of the same mathematical nature as the equipment layout, clusterisation and routing problems arising in corporative communication systems. The paper deals with two combinatorial optimisation problems that arise in the networked design of robotic cells: finding the optimal structure of the networked designing coalition (which is similar to choosing facilities and optimal connections in communication network design) and optimal sharing of available resources between the members of the web-connected coalition. A specific objective of the second problem is to save communication bandwidth which is achieved by an optimal combination of two basic tools of video information processing: (1) traditional telerobotics tools based on video cameras, and (2) using a virtual reality modeling language (VRML) which is the means to display robots and their behavior with animated computer graphics (which lead to a dramatic saving of the communication bandwidth since only coordinates have to be sent to the users and back, instead of video data). Two efficient numerical algorithms are proposed for solving specific set covering/set partitioning and PERT-type network scheduling problems that ensure optimal selection of the coalition structure and fair allocation of available material and information resources among coalition participants. The efficiency of the proposed algorithms are carefully investigated via computer simulation. Computational results indicate that the algorithms are effective in solving problems of a practical size. Commercially available packages are mainly based on interactive design that involves an experienced user for decision making, while the software provides sophisticated 3D geometrical modelling only and allows to estimate an acceptability of a user-proposed solution. Therefore, automatic generation of acceptable solutions and their Paretooptimisation is a challenging problem for CAD of industrial robotic systems. This paper focuses on combinatorial optimisation problems that arise in computeraided design of robotic cells. In contrast to other works devoted to general FMS design (Heragu, 1992; Stecke, 1983), it takes into account the essential features of the webbased robotic technology and concentrates on optimal selection of the appropriate tools. Selected aspects of these problems have been previously discussed by a number of authors. Cook and Han (1994) have used combinatorial optimisation algorithms for robot selection and work station assignment which is treated as a two-dimensional multi-type bin packing problem. Because of NP-hard nature of this problem, an efficient tree-phase heuristic algorithm has been developed. Souilah, Mecheri and Bennesroune (1996) have also investigated intra-cell layout design problem and have proposed a solution based on simulated annealing (Laarhoven and Aarts, 1987) and taboo search (Glover et al., 1993). However, in spite of the good computational properties, the previously developed algorithms deal with the simplified models of robotic cells and can not be directly applied in industrial robotic CAD systems, while the proposed approach relies on exact 3D models of the cells which include manipulators, welding guns, positioners, conveyors, workpiece, welding fixture, protective fences, etc. The proposed algorithm enables a user to generate a given number of the best decision sets and to choose a single solution in interactive exchange between the participants of the coalition using Paretoopimisation technique. The latter is described in details in (Pashevich, 1996). It should be noted, that the developed algorithm is rather sensitive to the price assignments. Our another algorithm is a further development of the generalized PERT scheduling algorithm suggested by Levner and Nemirovsky (1994). The developed combinatorial optimisation algorithms ensure solution of the specific set covering and set partitioning problems, which arise in CAD of industrial robotic systems and take into account the essential technological features. The efficiency of the proposed algorithms has been carefully investigated via computer simulation. Modern commercially available packages for the robotic computer-aided design (CAD) are mainly based on interactive design that requires an experienced user for the decision making while the software provides 3D geometrical models only. Creating the global CAD for robotic systems of the next generation based on co-ordination, multi-criterial optimisation and standardisation in the framework of the multi-national consortium is the challenging problem suggested and partially solved in the present study. The global Internet-based design of robots has the following advantages in comparison with the traditional robotic design: • up-to-date exchange of know-how and the relevant information; • a possibility to interactively solve hard decision-making problem simultaneously by many research teams; • increasing role of standardisation of robotic products and its components; • creating electronic libraries of multi-national and multi-disciplinary know-how: data bases, standards, technologies, technical solutions, theoretical design methods, etc. We intend to supply the packages with optimisation tools and customise them for the design of industrial robots at several plants in Europe, Japan and Israel.