A Generative Approach to the Development of Autonomous Robot Software

The integration of new or existing software components into established architectures and the ability to deal with heterogeneity are key requirements for middleware and development frameworks for robotic systems. This paper presents SPICA, a software development framework for communication infrastructures of autonomous mobile robots. Utilizing the model-driven software development paradigm, communication and data flow can be defined on an abstract level. For this purpose, domain-specific languages and tools are provided that allow specification and generation of module communication infrastructures for communication between modules along with primitives for data management. The high-level platform-independent specifications are automatically transformed into low-level platform and programming language-specific source code. We illustrate the applicability of our approach with an elaborate example describing the design of a soccer robot architecture that has proven its strength during RoboCup 2006. Our experiences have revealed that SPICA is advantageous for prototyping as well as for building high performance systems

1. Introduction

Software systems for autonomous robots consist of different kinds of software components. Sensors enable the robots to retrieve information from their environments. Actuators, in contrast, are used to interact with the environment. These components are normally located directly on the robot. They resemble device drivers which need to be tightly coupled with the software. In addition, robotic systems typically comprise components for world modeling, self localization, behavior reasoning, or path planning, which realize the artificial intelligence part of the system. These components can be located directly on the robot or realized as self-contained modules (processes) located on remote systems as well.