Introduction

This Research Package is devoted to studying, developing, and experimentally testing generic control architectures for intelligent machine systems. Control architecture is a system’s integrator in the sense that it combines all the other subsystems (e.g. motion systems, HRI, perception, and navigation) so that the tasks (goals) given by the user can be performed in a meaningful way within the boundaries given by the various constraints (internal/external, temporal, spatial).

Adaptive and learning elements should be included so as to improve the architecture’s performance in the future. Current architectures are more or less application-specific solutions that are hardly generic in any sense. In the near future the number of real-world intelligent machines systems will increase rapidly and thus there is a clear need for a basic framework for a control architecture design both in a single and a multi-machine case.

The de facto control architecture solution used nowadays is a hybrid deliberative/reactive architecture that combines reactive and deliberative (planning) elements. This type of architecture uses the best features from both approaches: the reactive element provides fast responses in a dynamic environment, whereas the deliberative part integrates a priori information and makes possible planning, especially in navigation-related issues.

The five elementary modules normally present in this type of control architecture are: mission planner, sequencer agent, behavioural (resource) manager, cartographer, and performance monitoring & problem-solving agent. These basic components will be defined, formulated, implemented, and tested with several different intelligent machines working alone or together with other machines. Furthermore, a SICL for intelligent machines will be developed to simplify and unify the interactions between the user and the machine.