The work presented in this paper deals with the development of a methodology

for resource-efficient behaviour synthesis on autonomous systems. In this context, a definition

of a maximal problem with respect to the resources of a given system is introduced. It

is elucidated by means of an exemplary implementation of the solution to such a problem

using the mini-robot Khepera as the experimental platform. The described task consists of

exploring an unknown and dynamically changing environment, collecting and transporting

objects, which are associated with light-sources, and navigating to a home-base. The critical

point is represented by the accumulated positioning errors in odometrical path-integration

due to slippage. Therefore, adaptive sensor calibration using a specific variant of Kohonen’s

algorithm is applied in two cases to extract symbolic, e.g. geometric, information from the

sub-symbolic sensor data, which is used to enhance position control by landmark mapping

and orientation. In order to successfully handle the arising complex interactions, a heterogeneous

control-architecture based on a parallel implementation of basic behaviours coupled

by a rule-based central unit is proposed.