4. Autonomous Dynamic Reconfiguration — Improving Collaborative Problem Solving

Abstract

Motivated by the pitfalls of collaborative problem solving in [121], [123], solution quality and communication efficiency, in this chapter [113], [117] we set forth some important requirements to avoid these pitfalls and inspect the literature for possible solutions. We state a lack of self-organization research on the individual level of agency rather than on the social level. On this basis, the AuReCon problem of finding a good configuration for a given distributed constraint optimization problem is formalized as a constraint optimization problem itself and proven to be at least NP-complete.

In the main section our concept to solve this problem is presented - the controlled usage of local reconfiguration operations called agent melting and agent splitting. In a detailed analysis we investigate the properties of this concept, namely its relation to the structure of the AuReCon problem, its impact on the underlying collaborative problem solving process and its sufficiency to solve the AuReCon problem. In addition, we investigate concurrency and autonomy issues in AuReCon.