Agenda

Opening Remarks (room 1060)

The Honorable Gary Peters
United States Senate

Michael Cadieux
Director, U.S. Army DEVCOM GVSC

Keynote:

COL Jeffrey Jurand
Project Manager, Maneuver Combat Systems, Program Executive Office Ground Combat Systems

Keynote:

Dr. Mario Santillo
Technical Leader, Robotics & Automation, Ford Motor Company

Break

Case Study:
The Artificial Bat: Pitting Echolocation-Like Ultrasound Perception Against Deployable Acoustic Cloaks

        Ultrasound-based perception systems replicating the biosonar capabilities of animals such as bats and marine mammals (Project #1.37) are effective at creating labeled maps of their surroundings in adverse weather. They are also significantly stealthier compared to the ubiquitous LIDAR- and radar-based systems and are relatively inexpensive while having low power consumption. In addition, since sound in air is mostly influenced by object geometry and less by material properties, it is significantly more difficult to develop countermeasures for sound-based perception methods (for instance, by judicious choice of target coating) than it is for electromagnetic-based systems such as LIDAR, radar, and video cameras. Nevertheless, sound-based countermeasures have been actively pursued. For example, cloaking structures capable of changing the objects' echoic signatures and be deployable on-demand in a small amount of time are already being investigated in the ARC (Project #3.21). This presentation highlights the lessons learned from pitting the perception system against the cloak and the possible improvements of these two adversaries.

Case Study:
Trailblazing Autonomy: Off-Road Adventures with AI Navigation

         Off-road navigation of autonomous vehicles presents far more stringent requirements than road terrains due to the vast variation of scenes caused by chaotic, unorganized, and natural geographical features and objects. Optimal autonomous navigation requires a perception system be accurate, precise, and require low computational effort to enable navigation and maneuverability at tactically relevant speeds. We have explored both image and lidar based semantic segmentation approaches to determine estimated trafficability of off-road environments which can be used to generate cost maps and occupancy grids to enable autonomous path planning.  After evaluation and comparison, an image-based convolutional neural network architecture was selected for segmentation and classification. To better optimize the perception system’s performance in the context of a relevant application, we integrated our perception solution with the NATO Autonomy Stack and performed test and evaluation iterations with the NATO AVT-341 Loyal Wingman Scenario. Iterations focused on strategies to (1) increase accuracy through composition of machine learning training datasets, (2) integration details to reduce computational effort, and (3) workflows for different sensor sets.

Lunch (atrium)

Technical Session 1 (matrix of parallel sessions)

Session 1.A (room 1060) and Session 1.B (room 1050)

Poster Session and Reception (atrium)
Reception begins at 4:30pm

Adjourn, Day 1

Day 2 Opening Remarks

Dr. Eric Michielssen
Associate Dean of Research and Louise Ganiard Professor of Engineering, University of Michigan

Dr. David Gorsich
Chief Scientist, U.S. Army Ground Vehicle Systems Center (GVSC)

Keynote:
NATO AVT-341: Off-Road Mobility Assessment Methods and Tools for Autonomous Military Ground Systems

Dr. Paramsothy Jayakumar
Senior Technical Expert, Analytics, U.S. Army GVSC

        Autonomous ground systems are a crucial part of the future military strategy and off-road mobility is a key measure of their performance. However, there is still a lack of understanding of, and metrics for the capabilities and reliability of these new technologies. How fast and reliably can a system reach its destination under a wide range of conditions? How well can it maneuver with soldiers under a variety of operations? Inability to answer such questions hinders fully fielding and operationalizing these systems. Therefore, the NATO Science and Technology Organization’s Advanced Vehicle Technology Panel has commissioned a Research Task Group, NATO AVT-341, to define methods and tools to assess off-road mobility of autonomous military ground systems.
        This talk presents the activities of the NATO AVT‐341, the achievements and challenges faced, development of fully functional modeling and simulation (M&S) platforms for evaluation of autonomous systems, and progress towards assessment of a militarily relevant scenario named Loyal Wingman. The M&S tools feature the latest physics‐based modeling approaches such as richly detailed off‐road environments, vehicles with multi‐body dynamics, NG-NRMM based vehicle‐terrain interactions, physics‐based exteroceptive sensor models, and modular ROS-based autonomy integration, along with a newly developed NATO Autonomy Stack, to simulate and evaluate autonomous operations in Loyal Wingman.

Panel Session:
Safely Navigating the Autonomy Landscape

Panelists:
Dr. Byron Boots
Amazon Professor of Machine Learning, Paul G. Allen School of Computer Science and Engineering, University of Washington

Dr. Cinzia Cirillo
Interim Director, Maryland Transportation Institute, Professor Civil and Environmental Engineering, University of Maryland

Dr. Matthew Johnson-Roberson
Director of the Robotics Institute, Professor of Computer Science, Carnegie Mellon University

Moderator:
Dr. Ram Vasudevan
Associate Director, Automotive Research Center

        Over the past year, the race to develop and deploy self-driving systems has been faced with critical challenges and several important questions loom large: How will the benefits of autonomous driving be shared with the public and deployed in the defense context? What regulatory mechanisms are needed to ensure the safety and reliability of these cutting-edge systems? What career opportunities will emerge in this space over the next five, ten, and twenty years? In the face of these challenges, this panel will take a deep dive into the current state of autonomy. Our experts will share insights on the latest technological advancements, explore the regulatory landscape, and examine what the job market in the autonomy space will look like. Join us as we navigate the road ahead and shed light on the future of self-driving vehicles.

Break

Technical Session 2 (matrix of parallel sessions)

Session 2.A (room 1060) and Session 2.B (room 1050)

Break

Awards & Lunch Reception (Atrium)

Announcing Winners of Best Student Poster Competition
Announcing Recipient of ARC Excellence in Research Award (ARC-ERA)
Celebration of Prof. Panos Papalambros's Induction to the National Academy of Engineering

One-on-One Info Session with GVSC Thrust Area Leaders (room 1060)

Dr. Jonathon Smereka, Thrust Area 1 Co-leader
Dr. Paramsothy Jayakumar, Thrust Area 1 Co-leader
Mr. Victor Paul, Thrust Area 2 Leader
Dr. Katherine Sebeck, Thrust Area 3 Leader
Dr. Vamshi Korivi, Thrust Area 4 Co-leader
Dr. Matthew Castanier, Thrust Area 5 Co-leader
Dr. Stephen Rapp, Thrust Area 5 Co-leader

Please email arc-event-inquiries@umich.edu to reserve a time-slot.

Access to and Use of the Robotic Technology Kernel (RTK) in ARC Projects (room 1050)
(additional registration required)

Dr. Jonathon Smereka
Senior Technical Expert, Ground Vehicle Robotics, U.S. Army GVSC

        RTK is an Army S&T autonomy software library of tested, vetted, managed, inter-compatible ROS packages which together establish a de facto common robotics platform and can be combined to form parts or all of an "autonomy kit/stack" for ground robots. This talk will introduce what RTK is, discuss the goals, and identify what is required for partners to work directly with the RTK software and material.

Day 2 Adjourns