Applying Evolution and Novelty Search to Enhance the Resilience of Autonomous SystemsNIER
We investigate the integration of evolutionary algorithms and novelty search in order to improve the performance and resilience of autonomous systems. We have developed two tools for this purpose: Evo-ROS and Enki. Evo-ROS combines evolutionary search with physics-based simulations of autonomous systems whose software infrastructure is based on the Robot Operating System (ROS). Enki uses novelty search to discover operational settings that lead to the most diverse behavior in the target system. Combining these tools yields an automated approach to explore the operational landscape of the target system, identify regions of poor performance, and evolve system parameters that better respond to adverse situations. In this paper, we present results of a case study of the throttle controller on AutoRally, a 1:5-scale autonomous vehicle designed by researchers at Georgia Tech for the study of aggressive autonomous driving. Preliminary experiments demonstrate the ability of the proposed methods to identify and characterize input speed signals that cause the existing controller to perform poorly. The ability to identify these troublesome signals enables development of a control system capable of handling a wider range of conditions by switching among controller modes that are optimized for different conditions.
Sat 25 MayDisplayed time zone: Eastern Time (US & Canada) change
14:00 - 15:30 | |||
14:00 25mTalk | Machine Learning Meets Quantitative Planning: Enabling Self-Adaptation in Autonomous RobotsLong Paper SEAMS 2019 Pooyan Jamshidi University of South Carolina, Javier Camara University of York, Bradley Schmerl Carnegie Mellon University, USA, Christian Kästner Carnegie Mellon University, David Garlan Carnegie Mellon University | ||
14:25 25mTalk | Self-Adaptation in Mobile Apps: a Systematic Literature StudyLong Paper SEAMS 2019 Eoin Grua Vrije Universiteit Amsterdam, Ivano Malavolta Vrije Universiteit Amsterdam, Patricia Lago Vrije Universiteit Amsterdam Pre-print Media Attached | ||
14:50 20mTalk | Applying Evolution and Novelty Search to Enhance the Resilience of Autonomous SystemsNIER SEAMS 2019 Michael Langford Michigan State University, Glen Simon Michigan State University, Philip McKinley Michigan State University, Betty H.C. Cheng Michigan State University | ||
15:10 20mTalk | Modelling and Analysing ResilientCyber-Physical SystemsNIER SEAMS 2019 Amel Bennaceur The Open University, Carlo Ghezzi Politecnico di Milano, Kenji Tei Waseda University / National Institute of Informatics, Japan, Timo Kehrer Humboldt-Universtität zu Berlin, Danny Weyns KU Leuven, Radu Calinescu University of York, UK, Schahram Dustdar TU Wien, Zhenjiang Hu National Institute of Informatics, Shinichi Honiden Waseda University / National Institute of Informatics, Japan, Fuyuki Ishikawa National Institute of Informatics, Zhi Jin Peking University, Jeffrey Kramer , Marin Litoiu York University, Canada, Michele Loreti University of Camerino, Gabriel A. Moreno Carnegie Mellon University, USA, Hausi Müller University of Victoria, Computer Science, Faculty of Engineering, Canada, Laura Nenzi University of Trieste, Bashar Nuseibeh The Open University (UK) & Lero (Ireland), Liliana Pasquale University College Dublin & Lero, Ireland, Wolfgang Reisig Humboldt-Universität zu Berlin, Germany, Heinz Schmidt RMIT Australia, Christos Tsigkanos Technische Universität Wien, Haiyan Zhao Peking University |