Publications

This paper characterizes the homotopy properties and the global topology of the space of positions of vehicles which are constrained to travel without intersecting on a network of paths. The space is determined by the number of vehicles and the network. Paths in the space correspond to simultaneous non-intersecting motions of all vehicles. The authors therefore focus on computing the homotopy type of the space, and show how to do so in the general case. Understanding the homotopy type of the space is the central issue in controlling the vehicles, as it gives a complete description of the distinct ways that vehicles may move safely on the network. The authors exhibit graphs, products of graphs, and amalgamations of products of graphs that are homotopy equivalent to the full configuration space, and are far simpler than might be expected. The results indicate how a control system for such a network of vehicles (such as a fleet of automatically guided vehicles guided by wires buried in a factory floor) may be implemented.

This report describes the results of composites fabrication research sponsored by the Laboratory Directed Research and Development (LDRD) program at Sandia National Laboratories. They have developed, prototyped, and demonstrated the feasibility of a novel robotic technique for rapid fabrication of composite structures. Its chief innovation is that, unlike all other available fabrication methods, it does not require a mold. Instead, the structure is built patch by patch, using a rapidly reconfigurable forming surface, and a robot to position the evolving part. Both of these components are programmable, so only the control software needs to be changed to produce a new shape. Hence it should be possible to automatically program the system to produce a shape directly from an electronic model of it. It is therefore likely that the method will enable faster and less expensive fabrication of composites.

We have developed, prototyped, and demonstrated the feasibility of a novel robotic technique for rapid fabrication of composite structures. Its chief innovation is that, unlike all other available fabrication methods, it does not require a mold. Instead, the structure is built patch by patch, using a rapidly reconfigurable forming surface, and a robot to position the evolving part. Both of these components are programmable, so only the control software needs to be changed to produce a new shape. Hence it should be possible to automatically program the system to produce a shape directly from an electronic model of it. It is therefore likely that the method will enable faster and less expensive fabrication of composites.

This report describes the results of assembly planning research under the LDRD. The assembly planning problem is that of finding a sequence of assembly operations, starting from individual parts, that will result in complete assembly of a device specified as a CAD model. The automated assembly programming problem is that of automatically producing a robot program that will carry out a given assembly sequence. Given solutions to both of these problems, it is possible to automatically program a robot to assemble a mechanical device given as a CAD data file. This report describes the current state of our solutions to both of these problems, and a software system called Archimedes 2 we have constructed to automate these solutions. Because Archimedes 2 can input CAD data in several standard formats, we have been able to test it on a number of industrial assembly models more complex than any before attempted by automated assembly planning systems, some having over 100 parts. A complete path from a CAD model to an automatically generated robot program for assembling the device represented by the CAD model has also been demonstrated.

We describe the implementation and performance of Archimedes 2, an integrated mechanical assembly planning system. Archimedes 2 includes two planners, two assembly sequence animation facilities, and an associated robotic workcell. Both planners use full 3 dimensional data. A rudimentary translator from high level assembly plans to control code for the robotic workcell has also been implemented. We can translate data from a commercial CAD system into input data for the system, which has allowed us to plan assembly sequences for many industrial assemblies. Archimedes 2 has been used to plan sequences for assemblies consisting of 5 to 109 parts. We have also successfully taken a CAD model of an assembly, produced an optimized assembly sequence for it, and translated the plan into robot code, which successfully assembles the device specified in the model.