The technologies of intelligent embedded systems, wireless sensor networks, and robotics at large, have dramatically developed in recent years, providing a wide variety of stand-alone systems capable of advanced individual performance. However, their very specialized nature might ultimately result in a limiting factor to their own spreading in mass applications. What is going to happen when, in a near predictable future, tens or hundreds of robots will move around in the same environment and share the same resources, each coming from
different industrial makers with completely different mechanical,
sensor, and software architectures, and each programmed to accomplish the goals of their owner? The design of a system of rules and protocols capable of regulating the robot behaviors so as to allow their co-existence as individually-driven agents and to avoid or limit the effects of conflicts, is the main objective of the first part of this work.
In engineering terms, the construction of the bases for a future
``society of robots'' to meet the specifications of real-world
applications, such as e.g. an automated robot highway or a shopping mall, entails addressing problems such as heterogeneity, scalability, safety, adaptability to environmental changes, and life-time. Enabling cooperation and integration of heterogeneous systems can help to address these non-functional requirements.
The first part of thesis presents the results of my research in the
field of controlling multi-agent systems. The aim of this work has been to implement a component-based control framework able to manage a large group of heterogeneous devices enabling cooperation. Only the assumption on wireless communication capability will be done for the devices, leaving out assumptions on other capabilities such as performances, dynamics, and specific tasks. A testbed based on three mobile vehicles, two of which where developed during this thesis, and a wireless sensor board running different operative systems was constructed and deployed in an application scenario.
Even though, enabling robot-to-robot cooperation can reduce the gap between robots and real application, we have to take into account that robots moving in the real world have to coexist and often cooperate with humans. The second part of this thesis will deal with this problem. The solution of the ``Safe Brachistocrone'', the optimal control problem with safety bounds, demonstrates the effectiveness of the Variable Impedance Approach (VIA) in guaranteeing safety and performance by adapting the mechanical impedance to the velocity of the link. From a control perspective a suboptimal solution able to be applied to general motion and computed in real-time is proposed and extended to be integrated with information on obstacle distance and velocity. Two prototypes of Variable Stiffness Actuators (VSA) will be showed, and the effectiveness of the proposed algorithms will be proved by experimental results.