A programming framework for perception-action systems

Abstract

In this paper we introduce CoolBot, a Component-Oriented programming framework for robotics. This framework’s main element is a language, which is presented at the specification level, and has been designed to assist robotic system developers in obtaining more structured and reusable systems without imposing any specific architecture. Within this framework components are conceived as Port-Based Automatons (PBA) that interact exchanging events and that can be composed to build up new components from existing components. Components, no matter if they are atomic or composite, are internally modeled as Discrete Event Systems and controlled using the same state control graph. CoolBot hides the programmer any aspects related to communications and provides standard mechanisms for different modes of data exchange between components. In the actual version of the system, CoolBot is used to declare the interfacing characteristics (number and type of input and output ports), parameters, internal variables, and a number of other aspects that affect the definition of the component as a PBA. This specification is then compiled to generate a C++ shell that the designer must complete supplying the required code to implement the component’s functionality. As a proof of concept we present two applications, an active vision system and a museum mobile robot, that have been developed on the basis of the ideas presented in this paper.