Abstract
The slow adoption of Artificial Intelligence (AI)-enabled agents that operate in dynamic real-world environments is in many instances due to the lack of model resilience and robustness to previously unseen real-world data, often leading to unpredictable behavior after deployment. This is true, for example, in commercial applications such as autonomous driving or human-robot interaction, and in military applications such as object recognition. It is therefore critical in many applications that the AI models generalize to new environment variations and scenarios. There are several approaches that one could consider, ranging from synthetic data augmentation approaches such as Sim2Real for robotics or incremental and continuous learning where an AI model dynamically adapts without forgetting what it has learned in the past. In this workshop, we seek novel advances in the development of AI models that are robust to distributional shifts in the data used to train AI models when deployed in dynamic environments. Emphasis will be placed on AI- models operating on multivariate time-series signals which are of relevance in signal processing.
Keywords: Artificial Intelligence, distributional shifts, continuous learning, decision support, autonomous systems, physics-informed models, Sim2RealOrganizers
Dr. Joel Goodman, Office of Naval Research (ONR), Arlingon, VA USA,
E: joel.i.goodman.civ@us.navy.mil
Dr. Shuowen Hu, DEVCOM Army Research Laboratory (ARL) Adelphi, MD USA,
E: shuowen.hu.civ@army.mil
Prof. Vikram Krishnamurthy, Cornell University, Ithaca NY USA,
E: vikramk@cornell.edu
Prof. Vishal Patel, Johns Hopkins University, Baltimore, MD USA,
E: vpatel36@jhu.edu
WebChair: Jay N. Paranjape, Johns Hopkins University, Baltimore, MD USA,
E: jparanj1@jhu.edu
Program Details
Duration: Full Day
Expected number of attendees: 50-100
Publication of the workshop papers at the IEEE Xplore Digital Library: Yes
Planned composition:
Overview (0.5 hr, by organizers)
Confirmed Keynote: Prof. Dinesh Manocha, University of Maryland, College Park (1 hr)
Panel discussion (1 hr)
2 Invited presentations (1 hr)
6-8 regular papers (3 - 4 hrs)
Poster Session (1- 2 hrs)
Diversity Inclusion and Ethical Issues
This workshop will promote an inclusive and equitable culture that welcomes, engages, and rewards all who contribute to the field, without regard to race, religion, gender, disability, age, national origin, sexual orientation, gender identity, or gender expression. We will work to ensure that any data or open-source material that our attendees have access to as a part of this workshop is free of explicit or implicit bias.