Human-Robot System Architectures
This study was based on a hypothetical future mission in which humans and robots explore the surface of Mars. We assumed a set of goals for the mission, decomposed the mission elements in terms of capability requirements, and overlaid the projected performance of human and machine agents to determine a suitable task decomposition.
Robots tend to perform as well or better than humans when the tasks and conditions can be reliably predicted. Humans are better-equipped to deal with the unexpected or abnormal. We compared performance under nominal and off-nominal conditions, and calculated the following results:
The 0.0 point represents the performance of autonomous robots. The red bars represent the performance of humans working without the aid of robots. We see that in the top section, which consists of nominal (normal, expected) situations, human performance ranges from inferior to slightly better than that of robots working alone. In the bottom off-nominal section, human performance ranges from slightly to far better than that of robots.
We analyzed four possible human-robot configurations for exploring a planetary surface, and displayed the results on the following graph, which maps performance vs. resources for the various human-robot systems.
This graph compares the resource demands and performance of two humans walking and working on Mars without robotic assistants with those of various combinations of humans working with robots. "Two humans walking" is indicated by the blue diamond in the center of the graph.
When performance rises faster or falls slower than demand on resources, compared to two humans walking, (+) value is added. If performance rises slower or falls faster than demand on resources, (-) value is added.
We see that the combination indicated by the diamond in the upper left quadrant (2 rover scouts operated by humans at a stationary command post) results in greater performance and less demand on resources than two humans on foot. The top diamond in the upper right quadrant (2 humans ride a rover) produces even higher performance, but requires more resources than two humans walking.
Please note that while this is the form in which we display different architecture characterizations for consideration by sponsors, the analysis for this particular study was performed in a first iteration before performance improvement projections and nominal/off-nominal considerations were introduced, and the integrated results were not recomputed.
For more information, contact: Charles.R.Weisbin@jpl.nasa.gov
Or see the following:
- "A New Method to Evaluate Human-Robot System Performance," (G. Rodriguez & C.R. Weisbin), Special Issue of Journal on Autonomous Robotics, Kluwer-Publishers, Vol. 14, pages 165-178, March 2003.