How Snappy Is Your Robot? Part 2 — Veo Robotics
What does the PSD look like as the robot stopping moves from snappy to sluggish? Figure 1 and Figure 2 show the calculated PSD parametrized for initial robot velocity v
r and robot deceleration a
s. As we can see, at sufficiently quick (but easily achievable) decelerations of 20-30m/sec
2
, the PSD is relatively insensitive to robot decelerations. In other words, the robot stopping time is only a factor in the PSD for slow-stopping robots.
The bigger question is which robot intrinsic parameter – robot controller latency T
r or robot deceleration a
s – has a larger impact on the PSD, and hence on the “collaborativeness” of the robot. To investigate this, we can parametrize the above equation in 3D space for a fixed robot velocity v
r. Figure 3 shows the 3D “envelope” for a robot velocity v
r or 2m/sec.
Clearly, the impact of the robot deceleration a
s is under 1 meter across the range of sensible robot decelerations (between 10 and 100m/sec
2
). On the other hand, for the range of observed robot controller latencies (50-500ms), the PSD range is around 2 meters.
Robot decelerations are a function of robot mechanics and kinematics, and are hard to make smaller, as physics (robot payload and inertia) get in the way. Fortunately, we find that easily achievable stopping performance is not a major factor in the PSD. The real culprit of large PSDs is robot controller latencies, which are a function of controller architecture and electronics. Designing low-latency controllers is conceptually possible but would require a rethinking of how robot companies design and develop controllers.
In the meantime, we at Veo Robotics are working on what is commercially available and find that there is a definite opportunity for human-robot collaboration with large industrial robots. In our next post, we will lay out the decisions end users need to consider when contemplating collaborative robotics applications, so stay tuned.
