After a very successful first project phase from 2016 to 2019, the LEAD project received a very positive midterm review by an international expert jury and is now entering a second phase from 2019 to 2021.
The underlying scientific goal of this project is to provide rigorous concepts, methods, and tools for the systematic construction of an Internet of Things that is resilient against failures due to adverse environments. We will offer methods and tools to predict, guarantee, and ultimately raise the level of dependability of the IoT. The project is structured around three broad research questions and three common threats that are tackled in three subprojects on dependable wireless communication and localization, verified security and real-time, and dependable multi-agent systems. Each of the subprojects combines the strengths of researchers from different domains and will be supported by two expertimental infrastructure: the Autonomous Diriving lab to investigate connect and autonomous cars such as platooning and the Automated Production lab which offers a collaborative robotic platform which recently won the RoboCup world championship in the Industrial League.
SUBPROJECT A "DEPENDABLE WIRELESS COMMUNICATION AND LOCALIZATION"
The use of Ultra Wide Band signals, in combination with adaptable transceiver frontends such as steerable, directive antennas – as investigated during the first project phase – has allowed us to increase the dependability of wireless communication and localization services. Most notably, we have achieved a reliable and highly accurate (centimeter-level) positioning despite the use of minimal infrastructure, which paves the way for new end-user applications. Accurate positioning also led to an improved location awareness: we have exploited locationresolved models of the environment to predict and control the dependability level of a wireless network. Building upon these findings, we now tackle the challenge of scaling up the developed system components towards realistic end-user scenarios. Consider, for example, large floor areas in industrial settings and retail (often highly cluttered by metal objects) and huge amounts of objects/products/nodes to be connected and located. We willspecifically investigate the scalability towards larger areas and more access points, more agent nodes, realistic, highly cluttered environments, and higher mobility of agent nodes. On the hardware side, we will investigate the impact of scaling the carrier frequency towards the mm-wave band. Millimeter-wave radios offer much higher bandwidth and a better beamforming capability, both of which can improve system-wide scalability and positioning performance. Increasing the system’s scalability will be the unifying, joint challenge addressed in the second project phase.
In Phase 1 we have developed learning-based techniques for testing the correct interaction in the IoT. By combining formal testing and verification with algorithms from AI, we detected several bugs in IoT protocol implementations as well as in our truck-lab demonstrator. We demonstrated that our methods work in practice and are currently transferring them to industry. In Phase 2, we will research novel formal design and development methods that can provide guarantees by construction. The aim is to automatically verify, at design-time and at run-time, concepts for security and Real-Time Operating Systems. We plan to build on the concepts that have been successfully developed in Phase 1 (SP 2) and also to research novel concepts that in particular focus on ease of verification. In order to achieve this goal, we approach verification as a triptych of techniques that complement each other: i) Formal verification: model checking and theorem proving provide the highest level of assurance at design-time; ii) Learning-based testing: where formal verification is not feasible, e.g., due to the lack of models, we will combine model-learning, testing and run-time verification; and iii) Run-time enforcement: since testing is incomplete, we will add fault-tolerant mechanisms that can detect misbehavior and trigger counter-measures at run-time. We will research formalizations of security and real-time concepts at an appropriate level of abstraction, taking nondeterminism, stochastics, and timing into account. New techniques will be developed that can learn, test, and enforce security and real-time requirements. We will also investigate the interplay of timing and security properties that are relevant in side-channel attacks.
In complex application scenarios of the Internet of Things, multi-agent systems embody a hierarchical control structure consisting of fast and reactive (e.g., servo position control) as well as a slower and deliberative control (e.g., mission planning and execution) in order to pursue complex long-term goals. This subproject addresses the dependability question in networked multi-agent control systems. To increase dependability in these systems, several innovations will be provided. Multi-agent systems need communication across networks and we will develop coding strategies that maintain closed loop stability under network impairments and furthermore we will characterize the quality of exchanged information for planning and execution monitoring purposes. Fault diagnosis and fault tolerant control are key techniques for dependable control systems. In a first step we use observer techniques to detect discrepancies between nominal and real system behavior. In a second step we interpret these discrepancies by exploiting either deterministic or probabilistic inference techniques. We address diagnosability and scalability issues in distributed environments. On cognition level, we develop distributed execution monitoring which is based on a cooperatively fused belief about the world. This belief state is enriched by the above mentioned inference techniques. Moreover, we investigate methods which allow to refine action models in the case of an execution failure. The refined action models will be used in the planning process to avoid future execution failures.
Wir verwenden Cookies, um Ihnen in Zukunft weiterhin das bestmögliche Service zur Verfügung stellen zu können. Nähere Informationen sowie Widerrufsmöglichkeiten finden Sie in der Datenschutzerklärung. Präferenzen speichern