RoSE Lab

Robotic Space Exploration

RoSE Lab’s Prinicpal Investigator is Dr. Frances Zhu and was started in January 2020. Dr. Zhu’s research revolves around exploring faraway, scientifically-rich environments, like the surface of Europa and Titan or the rings of Saturn, by applying Dr. Zhu’s dynamics and control knowledge and integrating machine learning models for better autonomy and adaptability. Dr. Zhu hope to develop these intelligent robots to autonomously navigate and probe scientific hotbeds in extreme terrains intermediately on Earth, like glaciers, hydrothermal vents, and underwater volcanoes, which offer analogues to space environments and independently have scientific value. Typically, scientists intuit the value of science and expeditioners intuit the danger of traversing extreme paths. An autonomous robotic scientist would instead digitize gathered measurements to scientific significance with proposed machine learning methods like Gaussian processes. A digital expeditioner would plan waypoints along a path, learn about an environment-vehicle dynamics model, and implementing low-level controllers on the vehicle to follow this path. A large gap in machine learning right now is the guarantee of safety for which I can derive for dynamic systems. The ultimate goal of Dr. Zhu’s research is to create algorithms that completely automate the planetary surface exploration process, digitizing the astronaut who would otherwise be in an immensely dangerous position. A future goal is to augment human exploration through cooperation and interaction with these intelligent robotic explorers. For more information on Dr. Zhu and her work please follow this link.

Active Learning Exploration Strategy

Characterize water ice distribution in lunar environment in situ.

Underlying kernel choice plays a key role in determining the performance of Gaussian Processes, a type of probabilistic machine learning that uses multivariate regression. My research proposes to find relationships between planetary surface characteristics and kernel choice to better improve the efficacy of surface exploration strategies built on Gaussian Processes.
Ashten Akemoto

Ashten Akemoto

System Identification of a Terramechanics Model

Accurately and adaptively predicts the interactions between the vehicle and the terrain in interpretable form, in real time, and in situ.

Currently developing a simulation for soft soil interaction with wheeled rovers. Utilizing deep learning with information from onboard sensors to analyze soil parameters and its affect on rover mobility during planetary exploration.

Bret Witt Jr.

Bret Witt Jr.

(Hoang) Long Ngyuyen

(Hoang) Long Ngyuyen

Bailey Hopkins

Bailey Hopkins

Simon Engler

Simon Engler

Simon is a PhD candidate from UHM that works along side the RoSE Lab to improve space exploration techniques with-in terra mechanics.

Celestial Navigation and Sensor Design

Precisely determine rover global position independent of orbiter.

The RoSE Startracker is uses celestial navigation to determine global position in GPS-denied environments. Using commercia-off-the-shelf (COTS) components and open-source software, the OpenStarNav algorithm is able to determine position in the global frame of reference via star images. The star tracker is a subsystem within the larger RoSE Lab autonomous planetary rover.

Baylor de los Reyes

Baylor de los Reyes

Kaiaka Kepa-Alama

Kaiaka Kepa-Alama