Dynamic Data Driven Pervasive Situational Awareness (DDDPSA)
collaborator: Sertac Karaman (MIT)
We are developing new adaptive planning techniques for pervasive battle-space situational awareness with ground and aerial sensors. This project is funded by AFOSR's DDDAS program.
Many modern Air Force missions take place in uncertain and dynamic environments, where a predictive model of the battlefield is essential to make effective decisions. These models are driven by data acquired using mobile sensors, such as optical and infrared cameras on board Unmanned Aerial Vehicles (UAVs). However, due to the enormous production costs and the presence of destructive adversarial behavior, it is impractical to deploy UAVs in large numbers. The emerging Unattended Ground Sensor (UGS) technology holds the potential to revolutionize Air Force operations in highly dynamic environments. UGSs can be produced cheaply and deployed in massive numbers. They can house seismic, acoustic, infrared, and optical sensors. However, the UGS may not have sufficient power to communicate their data to a central processing hub, necessitating data-ferrying with UAVs. This project seeks utilize and contribute to the develop Dynamic Data Driven Application System framework for developing UAV based data ferrying algorithms to facilitate battlespace picture development and situational awareness.
Our algorithm RAPTOR predicts which sensors are most likely to have interesting information.
AFOSR YIP: Robust Adaptive Autonomy in Contested Environments (RAACE)
We are developing new spatiotemporally scalable learning, planning, and manned-unmanned collaboration algorithms for teams of UAS operating in consented environments. This project is funded by AFOSR Young Investigator Program (YIP).
The research proposed in this YIP proposal seeks to develop theoretical underpinnings and practical algorithms for Robust Adaptive Autonomy in Contested Environments for mixed manned-unmanned aerial teams. Unmanned Aircraft (UA) have already seen deployment and success in diverse battle arenas, however, the current heavily-supervised UA operation paradigm is not well matched with the emerging needs of conflict. The proposed work includes the development of novel adaptive learning and decision-making algorithms that can provide robust mission performance in dynamically changing contested environments. The new approach pursued here departs from the emerging theory of Bayesian Nonparametric modeling, and leads to:
New scalable nonparametric predictive models and inference techniques for stochastic nonstationary processes with both long-term and abrupt changes;
New adaptive decision making algorithms that utilize these models for collaborative decision-making in uncertain, nonstationary, and contested environments.
The algorithms developed here have the potential to impact Air Force mission planning for manned-unmanned teams in the presence of threat, communication constraints, and dynamic adversaries. The fundamental limits and perforamnce guarantees of the developed algorithms will be mathematically characterized, and the algorithms will be carefully validated through simulations and flight-experimentaion.
We have developed kernel observers and Evolving Gaussian Process models for inference on spatiotemporally varying functions. See our NIPS paper on this work: Kernel Observers, Systems Theoretic Modeling and Inference of Spatiotemporally Evolving Processes
NRI: Collaborative Goal and Policy Learning from Human Operators of Construction Co-Robots
collaborators: Prabhakar Pagilla (TAMU),
Charles Abramson and Christopher Crick (OSU)
We are developing new methods to make it easier for novice operators to learn to use complex construction robots, such as excavators. This project is funded by National Science Foundation's National Robotics Initiative. NSF abstract and updates.
Terra-MEPP: High-throughput phenotyping for breeding better biofuel plants
TERRA-MEPP is a low-cost, autonomous robot that analyzes biofuel crops throughout the growing season. The robot’s sensors collect an unprecedented amount of field data, and high-throughput analytical strategies quickly analyze it to pinpoint plants with desirable yield and sustainability traits.
Past Sponsored Projcets
Surface and airborne sensors for detecting CO2 leakage
collaborator: Dr.Peter Clark, Dr. Jack Pashin, Dr. Tyler Lay, Dr. Jamey Jacob (OSU)
Our role in the CO2-Enhanced Oil Recovery Pilot in the Anadarko Basin (TX) is develop CO2 flux monitoring technologies using UAS and ground sensors. This project was funded by US Department of Energy (DOE).
Ensuring safe, permanent storage of CO2 in geologic carbon sinks is vital for the success of geologic storage projects. The National Energy Technology Laboratory of the U.S. Department of Energy has set a goal of 99 percent storage permanence in carbon capture, utilization, and storage (CCUS) projects. The development of monitoring technology that is capable of validating storage permanence while ensuring the integrity of CCUS operations is essential for meeting the goals of CO2 emissions reduction, environmental protection, and human health and safety. The identification of leakage pathways is the focus of this proposal. The proposed research program focuses on the design and deployment of a dense grid of shallow subsurface and surface sensors in combination with low-altitude airborne Unmanned Aerial Vehicle based detection of CO2 and CH4. These technologies will be deployed in the Farnsworth Oil Unit in the Anadarko Basin of the northeastern Texas panhandle.
Multi-Objective Optimization for Flexible Transport Aircraft
We created online adaptive multi-objective optimization based active wing-shaping controllers for next generation light-weight and flexible aircraft. This project was funded by NASA and OK-Space Grant consortium.
A Research Initiation Grant through NASA EPSCOR funded work in the multiple objective optimization (MO-Op) framework using a NASA model for a flexible wing Generic Transport (fGTM) aircraft based on the Boeing 757. Multiple, disparate objectives do not necessarily add together well to create a reward function for normal optimal control frameworks. Other methods are necessary to decide which objective is most important at a particular instant in order to optimize the operation. In some cases objectives are mutually exclusive such that meeting one means the system can not meet another objective. We showed that a subset of the problems at the forefront of this kind of control domains can be addressed with adaptive control and online learning.
DASLAB Experimental Infrastructure
DASLab Swarmy robot at the Chicago Field Museum, teaching kids about robotics and their field applications