Research Group
Computer Networks

We consider arbitrarily large systems (formally with an infinite number of nodes and from a practical point of view thousands of nodes), which are connected using technical channel models for mobile communication. The resulting graphs are random. An obvious question about such graphs is coherence (weak coherence component). However, in arbitrarily large systems, such coherence is too strict a conservative requirement, since a network node can always be isolated at times. Here, on the other hand, percolation as a weakened form of cohesion is a suitable system size. In infinite graphs, this means finding an infinitely large coherence component; in finite graphs, it is comparable to finding a very large coherence component that includes almost all network nodes.

We develop and study methods for structuring such random graphs, which are percolation-preserving and promote certain local structural properties of the resulting graph. Here we do not consider deterministic guarantees, but investigate to what extent such properties can be generated at a high rate. The properties themselves are subject to further investigation in terms of local communication and topology control methods.

The research project is funded in the context of the DFG project "Reactive construction of and reactive routing in Euclidean and topological network spans over wireless ad hoc and sensor networks".

The starting point of our research in this section is graphs with assumed local structure properties, such as can be observed or generated in high rate wireless networked systems. We design local methods under these assumptions, i.e., methods in which a network-wide goal is achieved based only on local decisions of the network nodes based on their neighborhood information. Arbitrary networks, which follow these structural assumptions, are structured with it. Graph coherence, spanner property, degree constraint and planarity are possible examples of such properties to be generated.

Another goal is to formalize the resulting graphs of such local methods using logical rules. Using such rules, we describe the results of the algorithms as so-called axiomatically defined graph classes. Many questions about whether our algorithms produce graphs with certain properties can be traced back to containment relations of graph classes. Using software tools from formal verification, we want to automatically derive guarantees for certain graph properties by means of such logically described graph classes and questions about containment relations formulated or programmed in such systems.

In this project, we consider cooperative control of mobile nodes (control of teams of AUVs or UAVs or fleets of vehicles, hereafter collectively referred to as agents) interacting over a wireless network. The network is established by the nodes themselves, i.e., no external communication infrastructure is assumed. Due to hardware limitations, in many cases only distributed approaches can be implemented. In such distributed cooperative control schemes, where each agent is equipped with a local control unit, the overall system performance is achieved by exchanging information over a wireless communication network. Due to the broadcast property of the wireless network, agents can send data from their local sensors to their immediate neighbors in one transmission. The subset of neighbors that further process the received samples in their local scheme is described by an interaction topology. This is a subgraph over the graph of all possible wireless communication channels. It is obvious that a change in the topology as well as error-prone communication will affect the performance and robustness of the interconnected system.

Cooperative control of multi-agent systems as well as topology control in wireless communication networks have been extensively studied in the literature, but so far separately. Results on cooperative control of multi-agent systems usually assume a given interaction topology. Research on topology control usually aims at improving the network in terms of communication properties. Since the interaction topology in a network of mobile agents has a significant impact on the achievable cooperative control performance, in this project we combine control schemes for dynamically controlling the movement of agents with adapting the interaction topology to the time-varying conditions of the communication network. Both simulatively and analytically, we demonstrate that such combined methods significantly improve achievable control performance (for example, concretely in terms of convergence rates).

The work in this project is funded by the DFG project "Cooperative Control and Topology Control in Wirelessly Networked Mobile Systems (AUVs and UAVs)" in the Priority Program SPP1914 "Cyber-physical Networking". The project is carried out in cooperation with the Institute of Control Engineering, TU-Hamburg-Harburg, the Center for Research in Electric Autonomous Transport (CREAT), Department of Electrical Engineering and Computer Science, University of Central Florida, and the Complex Systems Control Lab, School of Electrical and Computer Engineering, College of Engineering, University of Georgia.

This project focuses on 5G mobile communications for urban logistics and material transport in the factory environment using Automatic Guided Vehicles (AGVs). Here, the reliability of 5G mobile communication in the factory environment is a particular topic to be investigated. At a specific site, methods for 5G measurements are implemented in hardware and software in real operation. The goal is to derive statements about communication channels in the context of modern self-organizing factory systems. The question will be investigated to what extent ultra reliable low latency communication (URLLC) as defined in the 5G standard can be achieved in case studies and where the limits lie. Models will be derived for this purpose. Such models are crucial for the planning of future smart infrastructures such as smart cities and smart factories.

Furthermore, system integration issues will be considered, including the following: To what extent can technologies that are not part of the 5G standard be integrated? What fallback mechanisms can be found in the event that a 5G wireless connection degrades? To what extent can 5G be integrated with infrastructure-less communications, or to what extent can they be integrated with 5G?

This project is a co-supervision with Koblenz University of Applied Sciences as part of an industry-funded PhD project.

With the success of the Internet, communication behavior has changed significantly over the years: Communication participants are interested in content and less interested in needing to know the physical location of the content. Furthermore, in addition to communication behavior, the nature of access to networks has also changed: Subscribers are increasingly mobile (e.g., smartphones and laptops, mobile devices in the Internet of Things, connected vehicles). As a result, subscribers frequently change communication connections (switching networks, switching communication technologies (WLAN, mobile network)), which makes addressing and routing information to and from endpoints in a network, maintaining end-to-end connections, and identifying subscribers increasingly difficult.

In recent years, work has been done on technologies and concepts for the "Internet of the future", which aims to move away from host-based end-to-end communication towards data-centric communication. The principle: Instead of end nodes, data is addressed directly. This is summarized under the term "Information-Centric Networking" (ICN).

The constant networking of everyday objects is also finding its way into the automotive industry. Vehicles are increasingly equipped with sensors, actuators and communication units (mobile communications, WLAN, etc.). They form a mobile ad-hoc network, also called Vehicular Ad-hoc Network (VANet), and are thus able to exchange information with each other or with an infrastructure. However, vehicles in this network have a common denominator: they are highly mobile, making it difficult to exchange information using classical communication models.

In the context of this project, we will investigate to what extent concepts from data-centric networks can be transferred to the domain "Vehicular Ad-hoc Networks". Concepts such as addressing, caching and forwarding are investigated, new strategies are developed in this context and empirically evaluated by means of computer simulation and prototype implementation.

The research is carried out within the framework of an external doctorate with the central research department of Robert Bosch GmbH in Renningen.

In this project, we consider Software-Defined Networking (SDN) in the context of IoT with the goal of developing methods with the following characteristics. (1) Network management and scalability of wireless networks should be such that the entire network can be reprogrammed, configured and managed very easily. (2) Mobility of wireless sensor networks: When nodes move, the network topology changes. Consequently, time is needed to compute new paths (called routing convergence). Another goal of this project is to show how using the SDN concept will improve the above routing convergence and path quality. (3) Limited communication bandwidth, limited processing / storage capacity, and limited available power (from battery-powered nodes): The SDN approach will be designed to require only a small number of control messages, and solutions will be able to handle limited compute resources and meet different requirements of different applications (e.g., latency and throughput). In addition, in the case of battery-powered nodes, the SDN organization is designed to be very energy efficient.

Specific challenges in the project are: How to reach each node with its IP address? How to implement OpenFlow tables in nodes while reducing the need for OpenFlow control messages as much as possible? How to handle mobility and changing network topology to achieve largely transparent mobility / dynamic network change? How to keep the required number of control messages as small as possible, thus improving throughput and latency, improving power consumption and extending battery life?

Another research field worked on in our group is empirically investigated research questions using prototypes of cooperating flying robots. On a prototype of several flying Quadro copters with embedded Linux board more complex algorithms can be empirically investigated. Furthermore, a smaller swarm with about 30 micro UAV (CrazyFlies including extensions) is available in the lab.

Currently, we are dealing with questions of coordinated placement of robots, so that with the positions taken, the best possible network structure for the robot team is created with the maximum possible area covered. Here we deal with the design of algorithms based on mass-spring systems and grid-based structures and their evaluation in real experimental setups. Signal strength measurements (RSSI values) are used as manipulated variables for the mass-spring systems and grid structures.

Furthermore, we deal with the self-organized setup of support networks. This includes on the one hand the provision of a wireless network infrastructure by suitable positioning of the flying robots themselves, and on the other hand the dropping of relay nodes to suitable positions. For the latter, prototypical hardware implemented micro nodes are available. Suitable network formation protocols for very low-power microcontrollers are to be implemented on these nodes.