Evaluation and Complexity Analysis of Task Dependencies in an Artificial Hormone System

We study an extension of the Artifical Hormone System (AHS). The AHS is a decentralized task allocation system which has self-organizing properties. We extend the AHS by negator hormones where a task sending such a negator suppresses the execution of other tasks. We can realize conditional task structures in this way and use them to execute a task set on heterogeneous processors: If high-performance processors are available, high-performance tasks can be instantiated by the AHS. If the high-performance processors are failing, another task set running on general purpose processors is allocated (but with lower performance as before). We evaluate the negator approach by realizing two different control strategies for a self-balancing vehicle.The introduction of negators increases the complexity of deciding whether a communication path leading along several different tasks reaches its target eventually. We call this decision problem NEGATOR-PATH and prove its NP-completeness.