Supporting independent development, deployment and co-operation of autonomous objects in distributed control systems

Autonomous, active components like smart sensors and actuators offer the capabilities of spontaneous behaviour, concurrent computations and well-defined communication interfaces. The perspective of building a system from such active blocks however has an impact on modeling and deployment of the components and supporting their interaction at run-time. The paper presents a modeling approach for dynamically interacting autonomous objects addressing the problems of independent development, deployment and incremental extension. Compared to related approaches our concept further complements the modeling by an adequate middleware that supports the abstractions of the model during run-time and performs the respective dynamic binding and co-operation between objects. This opens a wide range of possibilities for dynamically integrating hardware and software components and support virtual sensors and actuators.

I. Introduction

Autonomous vehicles like cars and mobile robots are composed from a large number of smart networked components comprising hardware, software and, sometimes, mechanical components. These autonomous building blocks are equipped with a computational core and may exhibit spontaneous behaviour. Rather than just being transducers in terms of raw physical data, they offer the capabilities of information processing entities, actively providing application-oriented information via a network interface. Fig. 1 shows one of our robots which is equipped with eight motors, gyros and acceleration sensors, a compass, odometry sensors and distance sensors all equipped with their own processor and connected to the popular automotive CAN-Bus [1]. This forms the reactive system layer of the robot. Additionally, a more powerful computer performs the complex processing on the deliberative level and connects to a wireless network. In fact, the robot constitutes a distributed system and it would be highly desirable to purchase the components like navigation modules, environment perception equipment, or specific computing and signal processing engines from third party providers. Then, the programmer has just configuring an application along the information processing chain from sensors to actuators. The perspective of building a system from such active blocks however has a couple of consequences on modeling and deployment of the components and supporting their interaction at run-time. Figure 1. A network of smart components: The “Q” Robot