Interaction dynamics
An autonomous robot requires more than just geometric information about its environment, especially when it must engage in physical interaction with that environment: touching and grasping objects, pushing them around, assembling them, etc.
There are two major aspects to the concept of dynamical interaction with the environment, in that that environment can be
-
Passive: it has no sources of energy on its own, so only dissipates energy. The result is that the environment can never be the cause of instability when closing a feedback loop in the robot controller. However, stability of the robot controller is only one aspect of its success, the other being performance: it's not very useful to have a stable controller if that controller does not reach the goals specificied in the robot task…
-
Active: the environment is a controlled system in itself, e.g., another robot, or a human. This fact complicates the design of the robot controller to a large extent. The simplest approach is to make the robot controller become passive by itself, i.e., it reacts to active inputs from the environment by dissipating the energy generated in that interaction. While this is, in principle, a safe and stable approach, it severely limits the performance of the robot controller.