Comparative Analysis of Programming by Demonstration Methods: Kinesthetic Teaching vs Human Demonstration

TL;DR

Comparison of demonstration methods with comprehensive user study finds that human demonstration using a virtual marker is on average 8 times faster, superior in terms of quality and imposes 2 times less overall workload than kinesthetic teaching.

Abstract

Programming by demonstration (PbD) is a simple and efficient way to program robots without explicit robot programming. PbD enables unskilled operators to easily demonstrate and guide different robots to execute task. In this paper we present comparison of demonstration methods with comprehensive user study. Each participant had to demonstrate drawing simple pattern with human demonstration using virtual marker and kinesthetic teaching with robot manipulator. To evaluate differences between demonstration methods, we conducted user study with 24 participants which filled out NASA raw task load index (rTLX) and system usability scale (SUS). We also evaluated similarity of the executed trajectories to measure difference between demonstrated and ideal trajectory. We concluded study with finding that human demonstration using a virtual marker is on average 8 times faster, superior in terms of quality and imposes 2 times less overall workload than kinesthetic teaching.

Figures (7)

• Figure 1: The experimental setup for the user-experience study comprised a motion capture system, specially developed virtual markers, and the collaborative robot Franka Emika Panda. Participants were tasked to draw simple pattern using human demonstration with virtual marker (human operator on the left) and kinesthetic teaching with cobot (human operator on the right).
• Figure 2: NASA raw task load index (rTLX) questionnaire results. Bar charts show mean ratings of all participants. Line plots show frequency of certain rating for each task in the data acquired. For the workload measurement, lower rating is better. Graphs show that virtual marker introduces significantly less workload compared to the robot guidance.
• Figure 3: Left plot shows mean System usability score (SUS) (higher is better). Right plot shows saverage time for different programming modality (lower is better). Users determined that virtual marker that leaves trace is best in terms of usability. From the demonstration time perspective, using virtual marker speeds up demonstration process on average, eight times.
• Figure 4: Top: the recordings of a single participant with the cobot (no-trace in green and trace in yellow) and the virtual marker (no trace in blue and trace in red). The detail demonstrates the sampling of demonstration, with marked point $P_i$ on ideal line $l_i$ and point $P_d$ on demonstrated trajectory. Middle: the distance error between the demonstrated trajectories and the ideal trajectory for each segment (A-H). Bottom: contact force for each demonstrated trajectory along the drawing path.
• Figure 5: The mean value of all demonstrations (blue), compared to the ideal trajectory highlighted in red. The mean value is derived from distances between the points of trajectories and their respective orthogonal projections onto the ideal segment line. Yellow indicates the envelope of trajectory values, while the green area portrays the standard deviation from the mean trajectory.