WP1. Complex systems and interconnected objects
Understanding and controlling the complexity of systems of interconnected objects is a major challenge for both industrial and everyday life applications. We think, in particular, to fields like robotics, car industry, energy distribution or smart buildings, where it is essential to tackle autonomous heterogeneous objects and to develop robust control tools to optimize their interconnections. Our research in this direction will be developed within three interconnected tasks.
T1.1 Control, Optimization, Distributed Algorithms
François Clautiaux, IMB
This tasks aims developing new tools for the analysis, optimization and control of networks of heterogeneous systems.
The challenge is to develop control and optimization algorithms that are both precise and do scale-up to large distributed systems. Our aim is to contribute significantly to the understanding of these complex problems by carrying a mathematical analysis of “full models” (often involving partial differential equations), model reduction, deriving on-line control laws and using advanced optimization techniques to obtain highly efficient distributed algorithms. An important issue is to tackle uncertainties, which are inherent to the complexity of the considered systems, ensuring robustness of the planned solution or the control laws and developing tools for risk management. Regarding distributed algorithms, IMS and LaBRI have several contributions in the field of sensor network design. These networks are now ubiquitous, and collect and gather information in their environment. The teams involved in this work package have contributions at different levels: conception of the sensor network (sensors themselves, but also the structure of the network, the telecommunication used, and the middleware, which is used to produce an abstraction of the network and mask the heterogeneity of both the hardware and software parts of the sensors). These labs are also expert in self-organizing aspects of distributed algorithms, and signal processing and multivariate analysis of centralized or distributed data, depending on the size of the data to process. This clearly applies to robots and drone, which often have to make decisions autonomously, and of course sensor network design. The results and methods developed in this work package will form sound scientific foundations to share with the other methodological work packages and to be implemented in the transverse application-driven work packages.
T1.2 Large Scale Networks and Internet of Things (IoT)
Toufik AHMED, LaBRI
The IoT networks are composed of large number of nodes (sensors and actuators) that communicate with each other using wireless multihop connections to offer a set of various applications and services (motoring, data gathering, dissemination, etc.). A major challenge in developing IoT networks is the support of large number of diversified objects that use different types of radio interfaces and exhibit different requirements in terms of communication rages, channel interferences, etc. Nodes often operate on irreplaceable batteries with limited energy, transmission power, memory and computing capabilities. Furthermore, IoT network topology may be subject to frequent changes due to nodes' failure, nodes' mobility, in addition to some energy conservation mechanism used to adjust the transmission power or to schedule the node activity. To overcome these challenges, we aim in this work package to develop new solutions, mechanisms and protocols to support interoperable mass of IoT by facilitating generic model that can be used across verticals and beyond industry specific technologies. Toward this objective, we will interest in developing middleware for IoT interoperability, which abstracts physical network constraints, in proposition new routing protocols that facilitate communication, control and management of the network, in developing algorithms to reduce interference onto the network, and in optimizing the design of new sensors to offer flexible, reliable and low-cost specifications. The results of this work package is used as main foundation to build the 3 transverse pilots designed and deployed in WP5-WP7.
T1.3 Autonomous systems: interaction, cooperation and competition
Hugo GIMBERT, LaBRI
In this task we will develop tools for autonomous decision making in uncertain environment. The task T1.3 is strongly linked with WP7, which is devoted as well to autonomous systems, with a focus on perception of the environment and robustness of the control. The data gathered by embedded sensors is compiled in a model of the environment (WP7) which is then abstracted and processed in order for autonomous systems to take good high-level decisions, including cooperation and competition (T1.3). The main goal of task T1.3 is to produce methodological tools for high-level decision making in the context of cooperation and competition with other autonomous agents. We will especially study the case where the autonomous agents are bound to public communication and have to synchronize publicly without exchanging private messages, like is the case in Robocup competition. These tools will be tested on the robotic platform WP7, in several contexts including robotic competitions like Robocup (WP7).