Performance measurement of a mobile manipulator-on-a-cart and coordinate registration methods for manufacturing applications

Abstract

Mobile manipulators, which consist of a robotic manipulator arm on a vehicular base, enable more flexible and dynamic workflows in manufacturing processes by freeing the manipulator arm from the restriction of working on a single, calibrated workspace. A mobile manipulator-on-cart further enhances job concurrency since the manipulator arm is mounted to a detachable cart rather than directly to the vehicle payload structure. With this increase in flexibility, there is a need to understand new sources of position and orientation uncertainty. This uncertainty can degrade the accuracy and precision of the mobile manipulator position and orientation in addition to the coordination capability between the manipulator and vehicle. To facilitate performance evaluation and uncertainty characterization, the National Institute of Standards and Technology (NIST) has developed an artifact-based measurement methodology for mobile manipulator-on-a-cart systems. The artifact, which is re-configurable, simulates mock-assembly tasks for manufacturing applications and is compared to a ground truth consisting of an optical tracking system. In this paper, we document experimental procedures, data analysis methods, and results in measuring the performance of the manipulator-on-a-cart system. We also compare the speed and accuracy of two example coordinate registration techniques to be used by the mobile manipulator-on-a-cart to reduce position and orientation uncertainty.

Design and implementation of a closed-loop mobile manipulator control system

“…The same mobile manipulator-on-a-cart 4 featured in Ref. [26,29] and shown in Fig. 1 was used for this work.…”

“…As in previous experiments, the lower front sensor was disabled to prevent erroneous detection of the foor as an obstacle [26]. Additionally, the rear sensor, which was also not needed, was disabled and the maximum height at which the left and right side scanners could detect obstacles was kept at a reduced value (lowered to 250 mm from 1000 mm) [26,29]. The latter was done previously to prevent the CMMA from erroneously being detected as an obstacle [29].…”

“…Additionally, the rear sensor, which was also not needed, was disabled and the maximum height at which the left and right side scanners could detect obstacles was kept at a reduced value (lowered to 250 mm from 1000 mm) [26,29]. The latter was done previously to prevent the CMMA from erroneously being detected as an obstacle [29]. Under the path planning settings, the free space between the robot base and any obstacles along a planned path was increased from 300 mm to 700 mm to avoid collisions between the AMR-CT side laser scanners and the CMMA.…”

“…Finally, the deltaHeading and stop commands were used to allow for heading corrections in a manner similar to that implemented in Ref. [29].…”

“…The continuous performance scenario, which was evaluated using the Type B CMMA had been previously conducted in simulation environments and evaluated on a prototype mobile manipulator platform [27,28]. The most recent developments in measuring continuous mobile manipulator performance featured integration of the previously developed algorithms and simulations with the mobile manipulator-on-a-cart, which used a commercially oriented mobile platform and manipulator [29]. A more comprehensive summary of previous development milestones and evaluation activities can be found in Ref.…”