Skip to main content
Interactive Robotics Laboratory
Yu Gu, Professor

Nick Ohi

Decision Making Under Uncertainty
Aerospace Engineering


    BS Mechanical Engineering, West Virginia University (2016)

    BS Aerospace Engineering, West Virginia University (2016)

    PhD Aerospace Engineering, West Virginia University (EST 2022)

        Dissertation: “Improving Robotic Decision-Making in Unmodeled Situations”


Google Scholar


Nick Ohi is a native of Morgantown, West Virginia and is currently pursuing a Ph.D. in Aerospace Engineering at WVU. Nick supports autonomy and mission planning capabilities, systems integration, and field testing for numerous robotics research efforts (including precision robotic pollination and fast traverse for future NASA Mars rovers). While at WVU, Nick has worked as the lead programmer and software architect for the WVU team that won the NASA Sample Return Robot Centennial Challenge, he developed the flight software for the Global Positioning System - Precise Orbit Determination experiment flown on the Simulation-To-Flight 1 CubeSat mission, and he is the designer and maintainer of the WVU Robotics High Performance Computing Cluster.

Outside of his work at WVU, Nick runs the YouTube channel Coil Labs, which is based around his hobby of building Tesla coils and other projects involving high voltage electricity. Nick also volunteers as a mentor for the Mountaineer Area RoboticS (MARS) high school FIRST robotics team. In his spare time, Nick enjoys hiking, snow skiing, water skiing, and other outdoor activities.


Decision-making under uncertainty enables robots to operate autonomously and perform tasks to aid humans or to perform tasks that are tedious, dangerous, or even impossible for humans. Most realizations of autonomous robotic decision-making under uncertainty rely on accurate models of the decision-making problem, as well as models describing the distributions of uncertainty involved in the problem. If such models are not known, it is assumed that the parameters that define the models and distributions of uncertainty can be found through machine learning. Complex real-world situations, however, are likely to present unhandled edge cases, undefined situations, and unknown sources of uncertainty (i.e., model ambiguity) that violate the robot’s assumptions about the nature of the decision-making problem. Most existing methods are only demonstrated to work in controlled environments, such as warehouses and offices, where the environment can be altered to accommodate the robot’s limitations, thereby “eliminating the unknowns.” The inability of existing methods to generalize to more complex real-world situations, where unmodeled sources of uncertainty cannot be removed by altering the environment, is a key obstacle to bringing robots “out of the factory” and “into the real-world.”

Therefore, the objective of Nick’s research is to investigate autonomous robotic decision-making techniques for mobile robots that are “resilient to surprises,” where “surprises” are defined as violations of the decision-maker's assumptions about the nature of the decision-making problem. These violations of assumptions include: 1) additional or different states that define the true problem that are not included in the decision-maker's state space and 2) differences in the decision-maker's model of the decision-making problem, whether deterministic or stochastic (i.e., distributions of uncertainty), compared to the true dynamics of the real-world, either in structure, parameters, or both.


  • decision-making under uncertainty
  • autonomous systems
  • decision-making under ambiguity