Safety, Adaptability, and Security for Robust Autonomy of Robotic Systems
Presenter: Prof. Farshad Khorrami
NYU Tandon School of Engineering (Polytechnic Institute)
New York University
DEIB - 3B Room (Bld. 20)
December 20th, 2024 | 2.30 pm
Contact: Prof. Marco Mezzavilla
NYU Tandon School of Engineering (Polytechnic Institute)
New York University
DEIB - 3B Room (Bld. 20)
December 20th, 2024 | 2.30 pm
Contact: Prof. Marco Mezzavilla
Sommario
On December 20th, 2024 at 2.30 pm the seminar titled "Safety, Adaptability, and Security for Robust Autonomy of Robotic Systems" will take place at DEIB 3B Room (Building 20).
The development of autonomous unmanned vehicle technologies and their deployment involves several core challenges in vehicle design, sensor data processing, data fusion, localization, navigation, world modeling, obstacle avoidance, path planning, collaborative mission planning and formation maneuvering, distributed sensing and monitoring, and control. To perform any of these tasks, robotic platforms need to fuse in real-time proprioceptive and exteroceptive sensor data for environment perception, building world models, localization, and task planning. To achieve this, a host of mathematical as well as machine learning-based algorithms have been developed. This talk will focus on safety, adaptability, and security of machine-learning based approaches for various robotic platforms. Providing guaranteed performance/certificates for deep networks is a challenging area. Additionally, possible attacks such as adversarial perturbations or backdooring are other potential issues that need to be considered. Advanced attacks have shown that some of the earlier defenses are fragile and will not provide the required level of performance.
In this talk, we provide defenses in both white-box and black-box scenarios for backdoor attacks and applications to autonomous vehicles. Lastly, methods to alleviate such fragility of learning-based systems to adversarial perturbations will be presented based on generative adversarial learning-based techniques and control barrier functions. Additionally, an overview of the activities at the Control/Robotics Research Laboratory will be given with some experimental results.
The development of autonomous unmanned vehicle technologies and their deployment involves several core challenges in vehicle design, sensor data processing, data fusion, localization, navigation, world modeling, obstacle avoidance, path planning, collaborative mission planning and formation maneuvering, distributed sensing and monitoring, and control. To perform any of these tasks, robotic platforms need to fuse in real-time proprioceptive and exteroceptive sensor data for environment perception, building world models, localization, and task planning. To achieve this, a host of mathematical as well as machine learning-based algorithms have been developed. This talk will focus on safety, adaptability, and security of machine-learning based approaches for various robotic platforms. Providing guaranteed performance/certificates for deep networks is a challenging area. Additionally, possible attacks such as adversarial perturbations or backdooring are other potential issues that need to be considered. Advanced attacks have shown that some of the earlier defenses are fragile and will not provide the required level of performance.
In this talk, we provide defenses in both white-box and black-box scenarios for backdoor attacks and applications to autonomous vehicles. Lastly, methods to alleviate such fragility of learning-based systems to adversarial perturbations will be presented based on generative adversarial learning-based techniques and control barrier functions. Additionally, an overview of the activities at the Control/Robotics Research Laboratory will be given with some experimental results.
Biografia
Farshad Khorrami received his bachelor’s degrees in mathematics and electrical engineering in 1982 and 1984 respectively from The Ohio State University. He also received his master’s degree in mathematics and Ph.D. in Electrical Engineering in 1984 and 1988 from The Ohio State University. Dr. Khorrami is currently a professor of Electrical & Computer Engineering Department at NYU where he joined as an assistant professor in Sept. 1988. His research interests include system theory and nonlinear controls, robotics, machine learning, cyber physical system security, autonomous unmanned vehicles, embedded system security, and large-scale systems and decentralized control. Prof. Khorrami has published more than 360 refereed journal and conference papers in these areas. His book on “modeling and adaptive nonlinear control of electric motors” was published by Springer Verlag in 2003. He also has fifteen U.S. patents on novel smart micro-positioners and actuators, embedded system security, and wireless sensors and actuators. He has developed and directed the Control/Robotics Research Laboratory at Polytechnic University (Now NYU) and Co-Director of the Center for Artificial Intelligence and Robotics at NYU Abu Dhabi. His research has been supported by the Army Research Office, National Science Foundation, Office of Naval Research, DARPA, Dept. of Energy, Sandia National Laboratory, Army Research Laboratory, Air Force Research Laboratory, NASA, and several corporations. Prof. Khorrami has served as general chair and conference organizing committee member of several international conferences. He is also an IEEE Fellow. email: khorrami@nyu.edu
