Adaptation is a key component of many technical control systems in several machines, devices and networks. The design of adaptive systems has been intensively investigated by classical system theory for more than 50 years. However, today more than simple adaptation and control is required to tackle the challenges in many new technical applications.
Another challenge of the society in 2020 is urbanization and the transition to a low-carbon society. The infrastructure in cities needs to be evolved to a smart system as well. Due to multiple independently interacting participants (cars, pedestrians, public transport), traffic can be also considered as a random and nonlinear system. A smart traffic infrastructure should be able to predict the “states” of this system to avoid dangerous situations by assisting the driver or to improve the efficiency of the traffic to reduce traffic jams and pollution. The interaction in this highly dynamic and random environment requires real-time remote computing and inference based systems in different scenarios: from vehicle to vehicle (V2V) to vehicle to infrastructure (V2I).
To tackle the massive pollution caused by a growing population, a continuous observation of our environment is required. Massive sensor fields are needed to continuously sense the important information. Today, this massive data flow is still difficult to handle. Furthermore, due to the technical constraints in many applications, sensor motes will have extremely simple hardware resulting in very limited communication capabilities and intelligence at each mote. Consequently, sensing the important information with a dense sensor network with very simple sensor motes is a big research challenge.
Many solutions for the above-mentioned examples of applications rely on a reliable estimation of the states of highly nonlinear and dynamic random systems and finding the correct inferences. This leads to the central research questions in this field: How can we design more robust, distributed and smarter systems?