Learning Interacting Dynamic Systems with Prediction using Neural Ordinary Differential Equations

Abstract: 

Modeling Interacting Dynamic Systems is an important topic due to its many applications including autonomous driving and physical simulations. Many approaches model Interacting Dynamic Systems in temporal and relational dimensions. However, these approaches usually fail to learn the underlying continuous temporal dynamics, agent interactions and their dynamic adaptation explicitly. In this paper, we propose a Dynamic Data Driven approach in the form of an interacting system of ordinary differential equations (ISODE). Our approach uses the latent space of Neural ODEs to model continuous temporal dynamics by incorporating distance and interaction intensity into agent dynamic interaction modeling. In addition, we show how to control and update dynamically without retraining an agent's trajectory when obstacles and targets are introduced dynamically. Extensive experiments reveal that our ISODE DDDAS approach compares favorably with the state-of-the-art. We also show how an agent given sensing can dynamically avoid suddenly appearing obstacles and how to effectively control the agent motion by introducing attractors and repellers.

Bio:

Dimitris Metaxas is a Distinguished Professor in the Computer and Information Sciences Department at Rutgers University. He is directing the Center for Computational Biomedicine, Imaging and Modeling (CBIM) and the NSF University-Industry Collaboration Center CARTA with emphasis on real-time and scalable data analytics, AI, and machine learning methods with applications to computational biomedicine, and computer vision. Dr. Metaxas has been conducting research toward the development of novel methods and technology upon which AI, machine learning, physics-based modeling, computer vision, medical image analysis, and computer graphics can advance synergistically. Dr. Metaxas has published over 700 research articles in these areas and has graduated 65 PhD students, who occupy prestigious academic and industry positions. His research has been funded by NIH, NSF, AFOSR, ARO, DARPA, HSARPA, and the ONR. Dr. Metaxas work has received many best paper awards and he has 8 patents. He was awarded a Fulbright Fellowship in 1986, is a recipient of an NSF Research Initiation and Career awards, and an ONR YIP. He is a Fellow of the American Institute of Medical and Biological Engineers, a Fellow of IEEE and a Fellow of the MICCAI Society. He has been general chair of IEEE CVPR 2014, Program Chair of ICCV 2007, General Chair of ICCV 2011, FIMH 20011 and MICCAI 2008 and the Senior Program Chair for SCA 2007.

Formal Verification of Neural Networks in Autonomous Cyber-Physical Systems

Abstract:

The ongoing renaissance in artificial intelligence (AI) has led to the advent of data-driven machine learning (ML) methods deployed within components for sensing, actuation, and control in safety-critical cyber-physical systems (CPS). While such learning-enabled components (LECs) are enabling autonomy in systems like autonomous vehicles and robots, ensuring such components operate reliably in all scenarios is extraordinarily challenging, as demonstrated in part through recent accidents in semi-autonomous/autonomous CPS and by adversarial ML attacks. We will discuss formal methods for assuring specifications---mostly robustness and safety---in autonomous CPS and subcomponents thereof using our software tools NNV and Veritex, developed partly in a DARPA Assured Autonomy project. These tools have been evaluated in CPS development with multiple industry partners in automotive, aerospace, and robotics domains, and allow for formally analyzing neural networks and their usage in closed-loop systems. We will also discuss relevant ongoing community activities we help organize, such as the Verification of Neural Networks Competition (VNN-COMP) held with the International Conference on Computer-Aided Verification (CAV) the past few years, as well as the AI and Neural Network Control Systems (AINNCS) category of the hybrid systems verification competition (ARCH-COMP) also held the past few years. We will conclude with a discussion of future directions in the broader safe and trustworthy AI domain, such as in new projects verifying neural networks used in medical imaging analysis.

Bio:

Dr. Taylor T. Johnson, PE, is A. James and Alice B. Clark Foundation Chancellor Faculty Fellow and an Associate Professor of Computer Science (CS) in the School of Engineering (VUSE) at Vanderbilt University, where he directs the Verification and Validation for Intelligent and Trustworthy Autonomy Laboratory (VeriVITAL) and is a Senior Research Scientist in the Institute for Software Integrated Systems (ISIS). Dr. Johnson's research has been published in venues such as CAV, EMSOFT, FM, FORMATS, HSCC, ICSE, ICDM, ICCPS, NFM, RTSS, SEFM, STTT, TNNLS, UAI, among others. Dr. Johnson earned a PhD in Electrical and Computer Engineering (ECE) from the University of Illinois at Urbana-Champaign in 2013, where he worked in the Coordinated Science Laboratory with Prof. Sayan Mitra, and earlier earned an MSc in ECE at Illinois in 2010 and a BSEE from Rice University in 2008. Dr. Johnson is a 2022 recipient of the Best Artifact Evaluation Award at FORMATS, a 2018 and 2016 recipient of the Air Force Office of Scientific Research (AFOSR) Young Investigator Program (YIP) award, a 2016 recipient of the ACM Best Software Repeatability Award at HSCC, a 2015 recipient of the National Science Foundation (NSF) Computer and Information Science and Engineering (CISE) Research Initiation Initiative (CRII), and his group's research is or has recently been supported by AFOSR, ARO, AFRL, DARPA, Mathworks, NSA, NSF, NVIDIA, ONR, Toyota, and USDOT.