• Peter Liggesmeyer, Fraunhofer IESE
•  Rasmus Adler, Fraunhofer IESE 
• Richard Hawkins, University of York 

An autonomous system is capable of independently achieving a predefined goal in accordance with the demands of the current situation. In safety-critical applications, the operational situations may demand some actions from the system in order to keep risks at an acceptable level. This motivates the implementation of algorithms that estimate, assess and control risks during operation. In particular, the risk assessment at runtime is challenging as it implies moral decision making about acceptability of risks: “How safe is safe enough?”. However, it is also challenging to find a suitable notion of risk. IEC and IEC standards define the term “risk” differently following two “root” definitions: “combination of the probability of occurrence of harm, and the severity of that harm” and “effect of uncertainty on objectives”. The first definition is related to the way how integrity levels like SIL and ASIL are determined at design-time. In the session, we will discuss in how far existing design-time approaches can be adopted to implement an autonomous risk management at runtime. For instance, is it reasonable to implement algorithms that determine integrity levels at runtime?

• Detlev Richter, TüV SüD:  

  Digital twin-based hazard analysis at runtime for resilient production

• Simon Burton, Fraunhofer IKS:  

  Prerequisites for dynamic risk management 

• Patrik Feth, Sick AG:   

  Sensors for Dynamic Risk Assessment 

• Michael Woon, retrospect:  

  Being Certain of Uncertainty in Risk 

• Sebastian Steinhorst, Technical University of Munich, Germany
• Mohammad Hamad, Technical University of Munich, Germany 

Today's vehicles are increasingly connected and tomorrow's vehicles will be automated, autonomous, capable of sensing their environment and navigating through cities without human input. This comes at the cost of a new set of threats and cyber-attacks that can yield high recall costs, property loss, and even jeopardize human safety. In this session, three partners of the nIoVe H2020 EU project will present security challenges and solutions to improve future autonomous vehicles' security. The session starts with a short presentation about insights on the challenges and limitations of providing cyberthreat protections for public transport autonomous shuttles. The second talk discusses the need to make autonomous vehicles proactively able to react to intrusions and the challenges to achieving such a capability. The last talk addresses further cyber-security challenges that nIoVe aims to solve for autonomous vehicles. The session will continue with an open discussion to discuss all the introduced challenges and other related aspects. 

• Niels Nijdam, University of Geneva (UNIGE), Switzerland 

  The Perils of Cybersecurity in Connected Automated Vehicles

• Mohammad Hamad, Technical University of Munich (TUM), Germany 

  Toward a Multi-layer Intrusion Response System for Autonomous Vehicles

• Konstantinos Votis, Institute/Centre for Research and Technologies Hellas (CERTH/ITI), Greece 

  Cyber-security Solutions for Autonomous Vehicles

• Andy Pimentel (University of Amsterdam, Netherlands)
• Martina Maggio (University of Saarland, Germany) 

Due to the increasing performance demands of mission- and safety-critical Cyber Physical Systems (CPS), these systems exhibit a rapidly growing complexity, manifested by an increasing number of (distributed) computational cores and application components connected via complex networks. However, with the growing complexity and interconnectivity of these systems, the chances of hardware failures as well as disruptions due to cyber-attacks will also quickly increase. System adaptivity, for example in the form of dynamically remapping of application components to processing cores, represents a promising technique to handle this challenging scenario. In this session, we address the (consequences of the) idea of deploying runtime adaptivity to mission- and safety-critical CPS, yielding dynamically morphing systems, to establish robustness against computational hurdles, component failures, and cyber-attacks.

• Clemens Grelck (University of Amsterdam, Netherlands) 

  The TeamPlay Coordination Language for Dependable Systems

• Sasa Misailovic (University of Illinois at Urbana-Champaign, USA) 

  Programming Systems for Helping Developers Cope with Uncertainty

• Stefanos Skalistis (Raytheon Technologies, Ireland) 

  Certification challenges of adaptive avionics systems

• Samarjit Chakraborty (U North Carolina, Chapel Hill, USA)
• Petru Eles (Linköping University, SE)

Modern autonomous systems - such as autonomous vehicles or robots - consist of two major components: (a) the decision making unit, which is often made up of one or more feedback control loops, and (b) a perception unit that feeds the environmental state to the control unit and is made up of camera, radar and lidar sensors and their associated processing algorithms and infrastructure. While there has been a lot of work on the formal verification of the decision making (or the control) unit, the ultimate correctness of the autonomous system also heavily relies on the behavior of the perception unit. The verification of the correctness of the perception unit is however significantly more challenging and not much progress has been made here. This is because the algorithms used by perception units now increasingly rely on machine learning techniques (like deep neural networks) that run on a complex hardware made up CPU+accelerator platforms. The accelerators are made up of GPUs, TPUs and FPGAs. This combination of algorithmic + implementation platform complexity and heterogeneity currently makes it very difficult to provide either functional or timing correctness guarantees of the perception unit, while both of these guarantees are needed to ensure the correct functioning of the control loop and the overall autonomous system. This is a part of the overall challenge of verifying the correctness of autonomous systems.

• Qing Rao (BMW, Munich, Germany) 

  New Era in Autonomous Driving and the Role of IT - Will Traditional Carmakers Keep Pace?

• Soheil Samii (General Motors R&D, USA) 

  Dependable sensing system architecture for predictable perception in autonomous vehicles

• Deepak Shankar (Mirabilis Design, USA) 

  Design Tools for Predictable Hw/Sw Architectures for Autonomous Vehicles

• Cong Liu (UT Dallas, USA)  

  Towards Timing-Predictable & Robust Autonomy in Autonomous Embedded Systems

• Hamed Tabkhi (University of North Carolina at Charlotte, USA) 

  Toward AI-in-the-Loop Autonomous Safety System - Algorithmic and Timing Challenges

• M. Christakis (MPI SWS)
• H. Hermanns (U Saarland, Germany)

From autonomous vehicles to Industry 4.0, from smart homes to smart cities – cyber-physical technology increasingly participates in actions and decisions that affect humans. However, our understanding of how these applications interact and what is the cause of a specific automated decision is lagging far behind. This comes with a gradual loss in understanding.

The root cause of this problem is that contemporary systems do not have any built-in concepts to explicate their behaviour. They calculate and propagate outcomes of computations, but are not designed to provide explanations. They are not perspicuous.

The key to enable comprehension in a cyber-physical world is a science of perspicuous computing. This session will discuss the foundational, the industrial and the societal dimensions of the perspicuous computing challenge. It is organized by the Center for Perspicuous Computing – TRR 248 – a Collaborative Research Center funded by the German Research Foundation DFG.

• Introduction: “Enabling comprehension in a cyber-physical world with the human in the loop”
  Holger Hermanns, Universität des Saarlandes
• Panel: “Is industry or society in need for perspicuous computing? Both? Or neither?”
  • Panel Moderator: Christel Baier, Technische Universität Dresden
  • Panelists:
    • Bernd Finkbeiner

      CISPA Helmholtz Center for Information Security

    • Christof Fetzer

      Technische Universität Dresden

    • Raimund Dachselt

      Technische Universität Dresden

    • Prof. Rupak Majumdar

      Max Planck Institute for Software Systems

    • Dr. Lena Kästner

      Universität des Saarlandes, representing EIS

Highly automated vehicles need high performance HW/SW platforms to execute complex software systems for safety critical functions. This is a usually underestimated challenge when automation comes to production vehicles. The session starts with two short presentations of platform architectures that approach the resulting design quality and safety challenge with different methods. The session will continue with an open discussion of these and possibly other approaches.  

• W. Steiner, TTTech, Austria 

  MotionWise - A Brief Introduction and Outlook 

• D. Pangercic, APEX.AI, USA 

  Open-source and Developer Centric SW Platform for the New Breed of Vehicles 

Self-awareness principles promise to endow autonomous systems with high degrees of adaptivity and resilience, borrowing from an abundance of examples in biology and nature. However, the engineering of dependable and predictable autonomous systems pose significant challenges for explainability, testing, and bounding safe behaviors. This session begins with short presentations by academic and industry speakers on these topics, and is followed by an interactive discussion with the audience. The first presentation by Prof. Andreas Herkersdorf (TU Munich) discusses how transparent machine learning techniques can be coupled with self-awareness  to improve dependability in autonomous systems. The second presentation by Dr. Ahmed Nassar (Nvidia) addresses issues in training and testing of self-aware autonomous agents. The third presentation by Dr. Prakash Sarathy (Northrup Grumman) describes how to bound the emergent behavior of autonomous systems using a self-aware dataflow computing paradigm. The session is then followed by an open discussion between the audience and the speakers on research challenges and future directions at the intersection of self-awareness and autonomy. 

• A. Herkersdorf, TU Munich, Germany.   

  Transparent ML as a means of enhancing dependable autonomy 

• A. Nassar, Nvidia, USA. 

  Continuous Training and Testing of Autonomous Agents: The Road to Self-Awareness 

• P. Sarathy, Northrop Grumman, USA. 

  Self-aware dataflow computing for Bounded Behavior Assurance