Performance benchmarks are being built as a modular set of task-based tests. These reconfigurable tests incorporate small part insertions and fastening methods such as threading, snap fits, and gear meshing using standard components including screws, nuts, washers, gears and electrical connectors. The test methods leverage factors identified by Boothroyd-Dewhurst (B-D) design for assembly (DFA) studies. These studies have already identified and tabulated various important factors based on manual human performance in an assembly task. For instance, size and symmetry of parts, tool usage, fixturing, mechanical resistance, mechanical fastening processes, visual occlusion, and physical obstruction are all influential towards time-based human performance. Designing benchmarking tasks that efficiently sample this design space will greatly aid the assessment of a robotic system and quickly identify its strengths and weaknesses.
Another important aspect of performance measurement is providing confidence in the measured results. Consequently, multiple repetitions of a task are required to generate enough data for benchmarking comparisons. Moreover, the use of various statistical tests including test for correlation, distribution, variance, and mean will help identify significant comparative differences in performance data. Conducting these tests would also help reduce the number of false claims that may be issued regarding a robot’s level of performance.
The development of assembly-relevant metrics and tests is part of a NIST-led effort sponsored by the IEEE RAS Technical Committee on Robotic Hands Grasping and Manipulation (RHGM) with the goal of identifying key competencies and characteristics of robotic systems using a robust set of formalized evaluations and benchmarks. Ultimately, these benchmarks can help to match robotic hand capabilities to end-user needs as well as to help provide developers and researchers insight for improving their hardware and software designs.
The current prototype task boards are outlined below. Although shown in vertical orientations, the task boards can be placed horizontally or in other orientations that are representative of specific assembly requirements. Procedures for conducting the tests and capturing performance in a statistically significant manner are being developed.
Task Board #1
Competencies: peg insertions, gear meshing, electrical connector insertions, nut threading
This task board is designed to quantify a robot system’s capability for performing peg insertions, gear meshing, electrical connector insertions, and nut threading. Design factors include size, shape, location, and type of parts.
Instructions for reproducing the task board can be downloaded here .
Design files for laser cut board can be downloaded here.
STL files for 3D printing connector housings can be downloaded here.
CAD models for all components can be downloaded here.
STL models for all components can be downloaded here.
Kit tray for all components can be downloaded here.
Generalized test metrics, methods, and data analysis recommendations can be downloaded here.
A similar task board was used during the IROS 2017 Robotic Grasping and Manipulation Competition.
The details of the competition can be found here.