Efficient communication in autonomous robot software

Software for autonomous robots solving challenging tasks in research or application is becoming increasingly complex. System integration has to deal with various different functional components. To decouple those components from each other and to enable a modular and reuseable software architecture a robot middleware is typically used. But this intermediate layer introduces significant additional overhead during run-time. In this work a methodology is described to utilize specific application characteristics to improve communication efficiency between different robot software modules. By composing several components in a single thread memory copying or locking operations can be avoided, when data is exchanged between those parts. The optimization can be achieved without compromising the advantages of a communication layer. Still the modifications are transparent to the maybe already existing components. Experimental results in the scenario of autonomous soccer-playing humanoid robots are presented and exhibit remarkable reduction in communication overhead. Furthermore this approach can be implemented in or on-top of other communication layers.