Search

A biologically inspired mechanism for robot action selection, based on the vertebrate basal ganglia, has been previously presented (Prescott et al. 2006, Montes Gonzalez et al. 2000). In this model the task confronting the robot is decomposed into distinct behavioural modules that integrate information from multiple sensors and internal state to form ‘salience’ signals. These signals are provided as inputs to a computational model of the basal ganglia whose intrinsic processes cause the selection by disinhibition of a winning behaviour. This winner is then allowed access to the motor plant whilst losing behaviours are suppressed. In previous research we have focused on the development of this biomimetic selection architecture, and have therefore used behavioural modules that were hand-coded as algorithmic procedures. In the current article, we demonstrate the use of genetic algorithms and gradient–descent learning to automatically generate/tune some of the modules that generate the model behaviour.

A robot brain scheme has been implemented, based on and controlled by analog circuits and micro-controllers. The modules have been classified as: (a) world signal processing, (b) world signal relevance assessing, (c) pre-motor decisions, (d) motor behavior, and (e) planning and sequencing motor behaviors. There are two types of pre-motor processing: main and world signal tracking. Motor behavior includes all the final motion units. Each of these five classes roughly corresponds to areas in a vertebrate’s brain and proved to be an effective robotic brain scheme as they assist in the development of greater complexity in robotic brains and a means to compare different implementations. The scheme stresses the importance of motor behavior modules fed by pre-motor decision modules. The pre-motor decision modules aim the movement while the motor behavior module creates the behavior. Finally, the planning and sequencing modules are imperative implementations in a robotic brain.

In this article we describe a new robot control architecture on the basis of self-organization of self-inhibiting modules. The architecture can generate a complex behaviour repertoire. The repertoire can be performance-enhanced or increased by modular poly-functionality and/or by addition of new modules. This architecture is illustrated in a robot consisting of a car carrying an arm with a grasping tool. In the robot, each module drives either a joint motor or a pair of wheel motors. Every module estimates the distance from a sensor placed in the tool to a beacon. If the distance is smaller than a previously measured distance, the module drives its motor in the same direction of its prior movement. If the distance is larger, the next movement will be in the opposite direction; but, if the movement produces no significant change in distance, the module self-inhibits. A self-organization emerges: any module can be the next to take control of the motor activity of the robot once one module self-inhibits. A single module is active at a given time. The modules are implemented as computer procedures and their turn for participation scheduled by an endless program. The overall behaviour of the robot corresponds to a reaching attention behaviour. It is easily switched to a running-away attention behaviour by changing the sign of the same parameter in each module. The addition of a “sensor-gain attenuation reflex” module and of a “light-orientation reflex” module provides an increase of the behavioural attention repertoire and performance enhancement. Since scheduling a module does not necessarily produce its sustained intervention, the architecture of the “brain” is actually providing action induction rather than action selection.