My research lies at the intersection of machine learning and robotics, with a focus on approximate inference (Bayesian learning) and path planning. Currently, I am investigating sparse Gaussian processes to tackle critical issues in robotics. These include generating explainable DNN predictions, sensor placement, multi-robot informative path planning, and robot motion planning.
PhD in Computer Science, 2024
University of North Carolina at Charlotte
MSc in Computer Science, 2021
University of North Carolina at Charlotte
BSc in Computer Science, 2018
Wichita State University
This paper addresses multi-robot informative path planning (IPP) for environmental monitoring. The problem involves determining informative regions in the environment that should be visited by robots to gather the most information about the environment. We propose an efficient sparse Gaussian process-based approach that uses gradient descent to optimize paths in continuous environments. Our approach efficiently scales to both spatially and spatio-temporally correlated environments. Moreover, our approach can simultaneously optimize the informative paths while accounting for routing constraints, such as a distance budget and limits on the robot’s velocity and acceleration. Our approach can be used for IPP with both discrete and continuous sensing robots, with point and non-point field-of-view sensing shapes, and for both single and multi-robot IPP. We demonstrate that the proposed approach is fast and accurate on real-world data.
Methane, a harmful greenhouse gas, is prone to leak during extraction from oil wells. Therefore, we must monitor oil well leak rates to keep such emissions in check. However, most currently available approaches incur significant computational costs to generate informative data collection walks for mobile sensors and estimate leak rates. As such, they do not scale to large oil fields and are infeasible for real-time applications. We address these problems by deriving an efficient analytical approach to compute the leak rate distribution and Expected Entropy Reduction (EER) metric used for walk generation. Moreover, a faster variant of a submodular function maximization algorithm is introduced, along with a generalization of the algorithm to find informative data collection walks with arc routing constraints. Our simulation experiments demonstrate the approach’s validity and substantial computational gains.