Advanced Computer Scientist, Artificial Intelligence Center (AIC)
Vidyasagar Sadhu is an Advanced Computer Scientist at SRI’s Artificial Intelligence Center (AIC). At SRI, Dr. Sadhu has worked on both government and commercial projects including but not limited to DARPA RFMLS program (RF spectrum monitoring using reinforcement learning), DARPA Ditto/Succeed program (machine learning based surrogate models to speed design of military systems), generative modeling of novel images leveraging state-of-the-art GANs and diffusion models, predictive maintenance of industrial machines using multi-modal time-series sensor data, natural language based robotic navigation using knowledge graphs and deep reinforcement learning and graph attention network based multi-agent hierarchical reinforcement learning for multi-agent cooperation.
Prior to joining SRI, Dr. Sadhu interned at Honda Research Institute (HRI), San Jose, CA, where he has developed a novel multi-task learning framework for predicting/detecting anomalous/challenging driving situations using CAN bus scalar data and Microsoft Research, Bengaluru, India, where he developed deep neural network based auto-encoding techniques for predicting/detecting anomalous building energy consumption patterns.
Dr. Sadhu holds a PhD from Electrical and Computer Engineering department of Rutgers University, New Brunswick, NJ. His PhD thesis (under the guidance of Dr. Dario Pompili) is on real-time autonomic decision making under uncertain environments for UAV-based search-and-rescue missions. His research interests include applied Artificial Intelligence (AI)/Machine Learning (ML), with a special focus on Multi-Agent Deep Reinforcement Learning (MADRL), Multi-Modal Time Series Modeling and Deep Generative Modeling.
Dr. Sadhu is a co-inventor of two U.S. patents (Systems and Methods of Anomalous Event Detection and Safety – aware Route Recommendation System and Method ) and IEEE Senior Member. He has received several awards including National Science Foundation (NSF) Student Travel Grant, NSF Center for Science of Information (SoI) Student Research Project Grant, Rutgers ECE PhD Research Excellence Award, IEEE Best Demo Award and a co-recipient of the highly competitive Qualcomm Innovation Fellowship (QInF) Finalist Award. Dr. Sadhu holds a Bachelors and Masters degree (integrated 5-year program) in Electrical and Computer Engineering from Indian Institute of Technology (IIT), Madras, India [2007-2012].
Key Projects
- AiCorb: https://medium.com/dish/from-photo-to-building-a-collaboration-in-generative-model-architectural-design-7a86c226bfd5
- DARPA Ditto: https://www.sri.com/press/announcements/sri-international-receives-contract-on-ditto-project-under-darpa-ai-exploration-program/
- Predictive Maintenance: https://medium.com/dish/predictive-maintenance-for-industrial-internet-of-things-a22299f6a30d
Recent publications
-
Machine Learning Aided GPS-Denied Navigation Using Uncertainty Estimation through Deep Neural Networks
We describe and demonstrate a novel approach for generating accurate and interpretable uncertainty estimation for outputs from a DNN in real time.
-
Night-Time GPS-Denied Navigation and Situational Understanding Using Vision-Enhanced Low-Light Imager
In this presentation, we describe and demonstrate a novel vision-enhanced low-light imager system to provide GPS-denied navigation and ML-based visual scene understanding capabilities for both day and night operations.
-
Vision based Navigation using Cross-View Geo-registration for Outdoor Augmented Reality and Navigation Applications
In this work, we present a new vision-based cross-view geo-localization solution matching camera images to a 2D satellite/ overhead reference image database. We present solutions for both coarse search for…
-
Cross-View Visual Geo-Localization for Outdoor Augmented Reality
We address the problem of geo-pose estimation by cross-view matching of query ground images to a geo-referenced aerial satellite image database. Recently, neural network-based methods have shown state-of-the-art performance in…
-
Augmented Reality for Marine Fire Support Team Training
To provide FiSTs with the “sets and reps” required to develop and maintain proficiency, the Office of Naval Research 3D Warfighter Augmented Reality (3D WAR) program is developing an affordable…
-
Optimized Simultaneous Aided Target Detection and Imagery based Navigation in GPS-Denied Environments
We describe and demonstrate a comprehensive optimized vision-based real-time solution to provide SATIN capabilities for current and future UAS in GPS-denied environments.