Truss assembly and welding by Intelligent Precision Jigging Robots

Abstract

This paper describes an Intelligent Precision Jigging Robot (IPJR) prototype that enables the precise alignment and welding of titanium space telescope optical benches. The IPJR, equipped with µm accuracy sensors and actuators, worked in tandem with a lower precision remote controlled manipulator. The combined system assembled and welded a 2 m truss from stock titanium components. The calibration of the IPJR, and the difference between the predicted and the truss dimensions as-built, identified additional sources of error that should be addressed in the next generation of IPJRs in 2D and 3D.

Precise assembly of 3D truss structures using MLE-based error prediction and correction

“…A dedicated on-orbit experiment to assemble a space telescope optical bench will require hardware that is even more precise than the sensors and actuators that we used in Komendera et al (2014a). It will also make use of external measurement systems, such as precise cameras, to further improve the estimates.…”

“…However, to be conservative in how we interpreted the precisions of the actuator and the laser sensor, and to allow room for the aforementioned physical processes, we inflated the noises to s L = 8 mm and s M = 1 mm, and simulated MLE assembly trials on the sequence in Figure 7. It is obvious that each individual strut length can only be within 16 mm 95% of the time, meaning that the tolerance of 10 mm cannot be guaranteed with the hardware we used in Komendera et al (2014a). That said, with MLE in the assembly process, the average trace is 3.13 × 10 210 m 2 , making the mean error 17.7 mm as shown in Figure 11, only slightly worse than the 16 mm per strut error.…”

“…While we have previously demonstrated fully robotic external manipulation and onboard precise distance measurements on a two-dimensional test platform (Komendera et al, 2014a), for this experiment, a human external manipulator is required to attach the IPJRs to struts and to place each new cell in a coarse configuration. Once placed, the IPJRs refine the cell by using measurements of the distances between node balls, either measured by the IPJRs themselves or measured by an external sensor.…”

“…Quadrotor assembly is a recent and popular development; cubic truss structures are assembled in Lindsey and Kumar (2013), a quadrotor assembly agent uses reinforcement learning to learn to handle unmodeled situations in Barros dos Santos et al (2013), andJimenez-Cano et al (2013) demonstrates control with a manipulator arm attached. Other approaches include truss climbing and assembling robots (Detweiler Komendera et al (2014a) used Ultra-Motion linear actuators with 8mm step resolution, and had onboard length sensing using a KEYENCE laser sensor with 1mm resolution. The external manipulator shown at the left, known as the Lightweight Surface Manipulation System (LSMS), positioned the IPJR onto the site of the next cell to be assembled, and also positioned the IPJR over a strut canister to grasp new struts (not shown).…”

SpRoUTS (Space Robot Universal Truss System): Reversible Robotic Assembly of Deployable Truss Structures of Reconfigurable Length

“…When this method is applied to in-space assembly, these robots are usually called "satellite arms". [11], [12], [18] With this method, the separation of the assembler from the structure requires a global external positioning system in order to align the elements, adding to the complexity and limiting the scale of the final design to that of the positioning system [11], [12], [13].…”