Skip to 0 minutes and 2 secondsIn this video, we discuss the rocking motion of ERIC, and how that rock is reduced using accelerometers and so-called velocity feedback. So ERIC has two wheels, in contrast to our earlier robots, which also had a caster wheel. The battery pack and circuit board are held in place with a 3D-printed unit which also helps the motors. These pivot on the axis which supports the wheels, which means that when ERIC accelerates, IE when it speeds up or changes direction, the battery and circuit board rotate about this axis. This so-called rocking motion means, for instance, that the echo location sensors are not emitting a horizontal signal, so sometimes ERIC will see an obstacle, but sometimes the echo will go above it.
Skip to 0 minutes and 47 secondsIt would be much better if we could stop rock, which, of course, is a control problem. So we need to measure the angle of the circuit board, for which we use accelerometers. So let's see the effect of this rocking motion using first a simulation of ERIC, and then explain how we reduce these oscillations by talking about another system. Here, we have a web page which illustrates ERIC and its rocking motion, and if I scroll down to see the controls, if I reverse direction, we see oscillation. If we speed up or slow down, we get oscillations. If we change to a slope, you see oscillations. But if I put control on and then reverse, the oscillations are very much damped out.
Skip to 1 minute and 35 secondsThat happens also when you speed up or slow down. But how does that control work? Well, I shall explain that using another web page where we not only have ERIC, but we also have a different system which behaves in a very similar way. And again, I'll scroll down to see the actual controls. So the other system has a mass suspended from a spring from a ceiling. And if you apply a force to that mass, it moves downwards, like so, and you see it oscillate. Mass comes down, the spring is extended, the spring then pulls the mass back up, and we see the oscillations. And if there was no friction, this would carry on indefinitely.
Skip to 2 minutes and 16 secondsBut there is some friction, so the oscillations damp down. But if I add more friction, do the same thing, you see the oscillations are damped out very quickly. Well, that's similar to the whole situation with ERIC. If I reverse direction, you see the oscillations, put control on, the oscillations damp out. So doing that and doing that are very similar.
Skip to 2 minutes and 43 secondsWhy does it work? Well, let's explain that by looking at the systems from a cybernetic point of view. In both the mass-spring system and the ERIC, we apply a force, in this case, which makes the mass-spring oscillate, so the height goes up and down. Here, we apply acceleration to ERIC, which means the angle of the board goes, rocks around.
Skip to 3 minutes and 7 secondsTo control that, we put in so-called velocity feedback. Like we measure the height, and any change in that will be the velocity. And friction is proportional to velocity, so the friction is having the effect of dampening the oscillations. Or with ERIC, we measure the angle using the accelerometer, and the change in that angle is used to control ERIC and reduce the oscillations. And because velocity is change in position, this is known as velocity feedback control. So in summary, we've seen how, when ERIC accelerates, its circuit board rocks. This happens when it changes speed or direction, or goes on or off a slope.
Skip to 3 minutes and 52 secondsThe simulation shows that rocking motion, and also how it is reduced with control, achieved by detecting the change in the angle of the board. And this is analogous to a mass-spring system, whose oscillation is reduced by having friction. Have a look at the two simulations of ERIC and the mass-spring system. Also watch the videos showing ERIC with and without control.
Whereas our early robots had two wheels for driving it along and a caster wheel at the front for balance, ERIC just has two wheels, which means its circuit board, motors and batteries pivot around the axis between the two wheels.
Whenever ERIC changes velocity (speed or direction changes), the boards and batteries rotate, and if the change is too big, ERIC can go upside down when it turns itself off.
Another problem is that if the board is pointing upwards too much, the sensors may miss an object in front of the robot.
Stopping (or at least minimising) ERIC’s rock is a control problem - and as ever this requires a measurement. The fact that the board is at an angle can be detected using its accelerometer.
But how should ERIC react to this measurement?
When ERIC accelerates (its velocity changes), then immediately the board starts to rotate - so the angle can be used to adjust the velocity in the opposite direction. This affects ERIC’s speed, but only for a short time.
We can see this in the video which includes a simulation of ERIC on a web page and some film of ERIC with and without control of its rock.
© University of Reading