Breakout Session: Safety Assurance
Breakout Session: Safety Assurance Attendance: • High level of interest on the topic • ~ 70 on sign-up sheets (with more who joined the session) -
35 from private industries
-
14 from academia
-
15 from government agencies and non-profit organizations
Breakout Session: Safety Assurance Meeting Agenda (I): Jonas NILSSON, Volvo Car Corporation, Gothenburg; Dependability and Verification for Self-Driving Cars – The Drive Me Approach Naohisa HASHIMOTO, Nat. Institute of Advanced Industrial Science and Technology (AIST), Japan; Concerning safety assurance on Automated Vehicle -Results and discussion based on the projects in JapanWalther WACHENFELD, Technische Universität Darmstadt (Germany), Safety Assurance Based on an Objective Identification of Scenarios – One Approach of the PEGASUS-Project Lutz ECKSTEIN, RWTH Aachen University (Germany), Institute for Automotive Engineering (ika); Developing and Assessing Automated Driving
Breakout Session: Safety Assurance Meeting Agenda (II): Tim Allan WHEELER, Stanford University, Intelligent Systems Laboratory; Establishing Trust in Autonomous Vehicles – an Aerospace Perspective Nidhi KALRA, RAND Center for Decision Making under Uncertainty; Driving Autonomous Vehicles to Safety Marcos PILLADO, Applus IDIADA, Spain; Functional validation and performance assessment of automated truck platoons in controlled environments Michael WAGNER, Carnegie Mellon University; Challenges in Autonomous Vehicle Testing and Validation Andrew LACHER, Unmanned and Autonomous Systems Research Strategist, The MITRE Corporation; Applicability of Lessons Learned from Aviation Safety Management System for Automated Vehicles
Breakout Session: Safety Assurance Summary of Key Findings/Lessons Learned from Breakout Discussion, Panel I: 1. A safe and fail-operational vehicle implies a lot of redundancy. 2. Safety-related tasks must be clearly divided between driver and autopilot. 3. Strong interdependencies between levels of automation and safety assurance approach. 4. Scenario-based testing needs a documented and traceable way where the tests are derived from. 5. Databases and test methodology have to deployed step by step. Panel I discussion: • What is the role of the driver for the different level approaches? • How data for safety assurance can be collected commonly and shared for safety design?
Breakout Session: Safety Assurance
Summary of Key Findings/Lessons Learned from Breakout Discussion, Panel II: 1. A scientific, unified framework to optimize and evaluate the safety will lead to trust in automated driving. 2. We cannot wait for a perfect safe AD system. Instead we should start with reasonable safety in order to improve and start as soon as possible with the potential of saving lives. 3. For platooning as one use case for automated driving how tests can be executed. 4. (Machine Learning) AD (sub-)systems should be tested for robustness with the most challenging test cases. 5. From aviation we can learn more for safety management processes than from their methodology. Panel 2 discussion: • How learning systems could be tested and why monitoring of unknown situations might be useful? • How to communicate the fact that the AD could not be tested as perfect before introduction. • The worlds of aviation safety and road safety were compared concerning risk management and their financial resources for safety.
Breakout Session: Safety Assurance Recommended Action Items: • Open data bases for test scenarios • Establishing a safety management system (mimicking the model from aviation?) • Scientific accepted validation methodology • Standards, standards, standards, … for validation (not for function) • Honest discussion about safety expectation to the public Personal remark: • Having different automation levels under one umbrella “automated driving” does not help to avoid confusion even in expert discussions, not to mention consumers!!!
35 from private industries
-
14 from academia
-
15 from government agencies and non-profit organizations
Breakout Session: Safety Assurance Meeting Agenda (I): Jonas NILSSON, Volvo Car Corporation, Gothenburg; Dependability and Verification for Self-Driving Cars – The Drive Me Approach Naohisa HASHIMOTO, Nat. Institute of Advanced Industrial Science and Technology (AIST), Japan; Concerning safety assurance on Automated Vehicle -Results and discussion based on the projects in JapanWalther WACHENFELD, Technische Universität Darmstadt (Germany), Safety Assurance Based on an Objective Identification of Scenarios – One Approach of the PEGASUS-Project Lutz ECKSTEIN, RWTH Aachen University (Germany), Institute for Automotive Engineering (ika); Developing and Assessing Automated Driving
Breakout Session: Safety Assurance Meeting Agenda (II): Tim Allan WHEELER, Stanford University, Intelligent Systems Laboratory; Establishing Trust in Autonomous Vehicles – an Aerospace Perspective Nidhi KALRA, RAND Center for Decision Making under Uncertainty; Driving Autonomous Vehicles to Safety Marcos PILLADO, Applus IDIADA, Spain; Functional validation and performance assessment of automated truck platoons in controlled environments Michael WAGNER, Carnegie Mellon University; Challenges in Autonomous Vehicle Testing and Validation Andrew LACHER, Unmanned and Autonomous Systems Research Strategist, The MITRE Corporation; Applicability of Lessons Learned from Aviation Safety Management System for Automated Vehicles
Breakout Session: Safety Assurance Summary of Key Findings/Lessons Learned from Breakout Discussion, Panel I: 1. A safe and fail-operational vehicle implies a lot of redundancy. 2. Safety-related tasks must be clearly divided between driver and autopilot. 3. Strong interdependencies between levels of automation and safety assurance approach. 4. Scenario-based testing needs a documented and traceable way where the tests are derived from. 5. Databases and test methodology have to deployed step by step. Panel I discussion: • What is the role of the driver for the different level approaches? • How data for safety assurance can be collected commonly and shared for safety design?
Breakout Session: Safety Assurance
Summary of Key Findings/Lessons Learned from Breakout Discussion, Panel II: 1. A scientific, unified framework to optimize and evaluate the safety will lead to trust in automated driving. 2. We cannot wait for a perfect safe AD system. Instead we should start with reasonable safety in order to improve and start as soon as possible with the potential of saving lives. 3. For platooning as one use case for automated driving how tests can be executed. 4. (Machine Learning) AD (sub-)systems should be tested for robustness with the most challenging test cases. 5. From aviation we can learn more for safety management processes than from their methodology. Panel 2 discussion: • How learning systems could be tested and why monitoring of unknown situations might be useful? • How to communicate the fact that the AD could not be tested as perfect before introduction. • The worlds of aviation safety and road safety were compared concerning risk management and their financial resources for safety.
Breakout Session: Safety Assurance Recommended Action Items: • Open data bases for test scenarios • Establishing a safety management system (mimicking the model from aviation?) • Scientific accepted validation methodology • Standards, standards, standards, … for validation (not for function) • Honest discussion about safety expectation to the public Personal remark: • Having different automation levels under one umbrella “automated driving” does not help to avoid confusion even in expert discussions, not to mention consumers!!!
