Dynamic stabilisation is the process of actively stabilising an inherently unstable object such as the Inverted Pendulum which will be covered in the next tutorial. This requires a real time control system which is in two parts. In the first part, we have some input signal describing the state of the object and in the second part, we generate a response to that signal to achieve a desired state.
When the system is very time sensitive as is the case with a falling pendulum, it is essential to operate without latency, that is, with as little delay as possible. This input signal also needs to be smooth because we don’t want our rapidly responding system to be responding to a noisy input signal that is merely a property of the sensor electronics. Unfortunately the requirement for low latency means we can’t rely on a simple rolling average to smooth the data. This simply won’t work with the T-Bot (you can try). Methods for for producing a clean signal will be discussed in the Kalman Filter tutorial.
Once we have achieved a nice smooth signal, we want to generate a measured or proportional response to compensate for the difference between our desired state and our measured state (we call this the error). In other words, if our error is small, we want our response to be small. And conversely, if our error is large, we want our response to be large. This is most commonly done with a PID (Proportional Integral Derivative) controller which will be discussed in the PID tutorial.