Emergent Control

Abstract

In order to control the dynamics of a system, feedback control (FC for short) is an extremely successful strategy, which is widely applied by engineers. Here we discuss a different strategy of control, called emergent control (EC for short), which can be found in large, distributed systems of components interacting only locally. For comparison we present a basic architecture for emergent control and two simple examples. In these examples, emergent control is achieved by a chemical computing approach. In the first example the number of objects of a particular type in a distributed system has to be kept constant. The example shows that on a macroscopic level EC and FC can display exactly the same behaviour. Hence for comparing their performance quantitatively a more refined model has to be taken into account. This model indicates a trade-off between cost and robustness. FC tends to operate at a lower cost than EC, however it also tends to instability when the system under control is large, decentralised, and/or heavily perturbed. In the second example the number of clusters in a distributed system should be controlled. The example shows how a user can “control”, i.e., provide goals in EC even if the system is not tractable analytically due to highly non-linear effect