Unmanned and Swarm Systems Lab

Our lab currently focuses on two research thrusts: a) control, decision making, and optimization for autonomous and sensing systems; b) application of artificial intelligence and computer vision for microscopy image analysis. The following are on-going projects:

Swarm Dynamics and Decision Making

We are developing decentralized decision making methods and information sharing schemes to produce cooperative behavior from a swarm of agents (e.g., UAVs). As most decision making problems lead to hard optimization problems, we investigate fast heuristic approaches and trade off optimality. Specifically, we are currently looking at the following UAV swarm control applications: formation flying, 3-D object shape reconstruction with a swarm of UAVs carrying 2-D sensors, and for precision agriculture. This research is currently funded by a seed grant from SD-GOED.

UAVs tracking targets
Left image: Visiting a farmland in Sturgis, SD for data collection via an agricultural drone (Abir (left, holding DJI Mavic 2 Pro drone), Ragi (right)). Right image: Image of the above farmland collected via a drone equipped with multispectral camera showing NDVI values.

Decentralized Decision Making

Fast decentralized decision making methods are the key in bringing real-time cooperative behavior in systems with multiple and independent decision makers or "agents". We are investigating group dynamics in decentralized systems, and developing decision making methods and information sharing schemes for agents in a networked (time-varying) system of agents. We primarily use decision theory, specifically dynamic programming principles, to develop these methods, and currently focused on developing approximate dynamic programming methods for decentralized decision making. Some aspects of this research are currently funded by an AFOSR grant.

Deep Learning and Computer Vision Methods for Microscopy Image Analysis

We are developing deep learning methods to analyze microscale and nanoscale image data for understanding rules of life in biofilms and 2D materials. Our goal is to predict the biofilm phenotype given the surface and 2D coating characteristics and the environmental factors. This research is currently funded by an NSF EPSCoR grant.

Segmentation of bacterials cells in a SRB-biofilm via deep learning