Publications
Please, if you have questions or would like a copy of a manuscript, email me or message me through the ‘About’ page.
Portable Bevameter Design for Geotechnical Characterization on Planetary Surfaces
Saito, J., Wang, H., Zhu, F., “Portable Bevameter Design for Geotechnical Characterization on Planetary Surfaces”, accepted to International Space Robotics Conference 2024.
The Artemis program’s plan to establish scientific lunar habitats has underscored the critical need to characterize material composition and geotechnical properties at potential landing and building sites. This paper assesses a crucial geotechnical parameter, sinkage, which is traditionally measured using bulky and massive equipment confined to laboratory settings. Current practices involve transporting field samples for controlled testing, often lacking assessment of sinkage in lunar regolith and simulants. To bridge these gaps, a novel portable bevameter design is proposed, operable by a single operator in the field. The accompanying procedure outlines the operation of the bevameter in situ and processing of raw data into geotechnical properties. The implementation has been validated against the NASA COLDArm measurements, with the sinkage parameters falling within the 95% confidence intervals of reference data. Additionally, the geotechnical characterization of lunar simulants LHS-1 and real analogue terrain samples is presented. This innovative design holds significant potential for advancing technologies crucial to both robotic and crewed lunar missions. Applications include in-situ terramechanics characterization affecting rover dynamics and mobile geotechnical surveying for habitat foundations, launch and landing pads, and large vehicle maneuverability.
Planetary Localization in GPS-Deprived Environments with Open-Source Software and Commercial-Off–the-Shelf Components
Kepa-Alama, K., de los Reyes, B., Tennebaum, A., Zhu, F., “Planetary Localization in GPS-Deprived Environments with Open-Source Software and Commercial-Off–the-Shelf Components”, accepted to International Space Robotics Conference 2024.
Autonomous unmanned vehicles can perform detailed surveys of planetary surfaces but it is imperative to establish a method for global localization to effectively explore these areas. The absence of GPS in extraplanetary environments prevents surface vehicles from knowing their exact location, which raises the need for an alternative positioning system. Traditional methods, such as visual odometry cross-referenced with digital elevation maps, are limited by their dependence on human input and pre-existing space infrastructure. This paper outlines and characterizes a global position determination algorithm intended for planetary surface vehicles in GPS-denied environments without any prior knowledge. The localization algorithm receives images of the stars from a visible camera and tilt measurements from an inclinometer, derives star locations and a gravity vector, and combines these signals to generate an onboard position determinant. This paper contributes (i) the first open-source planetary localization algorithm, (ii) a sensor suite design derived of solely commercial-off-the-shelf (COTS) components, and (iii) an unprecedented physical experiment and characterization of this LIS algorithm on Earth’s surface. The resulting position determinant is on average 100km from the testing location, consistent across various time and surface inclines. This achieved determinant error offers a starting point toward localization improvement with more capable sensors, more precise clocks, and the incorporation of multiple historical determinants for state estimation. Independent and precise planetary surface localization enables more autonomous, ambitious robotic missions.
Jitter Analysis of the HyTI Satellite
The Hyperspectral Thermal Imager (HyTI) is a technology demonstration mission that will obtain high spatial, spectral, and temporal resolution long-wave infrared images of Earth’s surface from a 6U cubesat. HyTI science requires that the pointing accuracy of the optical axis shall not exceed 0.014 mrad (approximately 2.89 arcseconds) over the 0.5 ms integration time due to these effects (known as jitter). Two sources of vibration are a cryocooler that is added to maintain the detector at 68 K to achieve acceptable dark current levels and three orthogonally placed reaction wheels that are a part of the attitude control system. Both of these parts will introduce vibrations that get propagated through to the satellite structure while imaging. Typical methods of characterizing and measuring jitter involve complex finite element methods, computationally expensive modeling, expensive equipment and specialized laboratory setups. In this paper, we describe a novel method of characterizing jitter for small satellite systems that is low-cost, simple, and minimally modifies the subject’s mass distribution. The metrology instrument is comprised of a fiber-coupled laser source, a small mirror that is mounted via a 3-D printed clamp to a jig, and a lateral effect position-sensing detector. The position-sensing detector samples at a Nyquist frequency of 500 Hz and can measure displacements as little as 0.15” at distances of one meter. This paper provides an experimental procedure that incrementally analyzes vibratory sources to establish causal relationships between sources and the vibratory modes they create. We demonstrate the capabilities of this metrology system and testing procedure on HyTI, using the advanced Attitude Determination, Control, and Sensing (ADCS) Test Facility in the Hawaii Space Flight Lab’s clean room. Results include power spectral density plots that show fundamental and higher-order vibratory modal frequencies in HyTI with a precision of better than one arcsecond measured at distances of approximately one meter. The metrology instrument and procedure can attribute correlation and possibly causation of these modal frequencies to vibratory sources. Results from metrology show that jitter from reaction wheels meets HyTI system requirements within 3σ.
Urasaki, C., Zhu, F., Bottom, M., Nunes, M., Walker, A. “Jitter Analysis of the HyTI Satellite”, 2024 IEEE Aerospace Conference Proceedings.
Akins, Sapphira, and Frances Zhu "Comparing Active Learning Performance Driven by Gaussian Processes or Bayesian Neural Networks for Constrained Trajectory Exploration." ASCEND 2023. 2023. 4720.
Robots with increasing autonomy progress our space exploration capabilities, particularly for in-situ exploration and sampling to stand in for human explorers. Currently, humans drive robots to meet scientific objectives, but depending on the robot’s location, the exchange of information and driving commands between the human operator and robot may cause undue delays in mission fulfillment. An autonomous robot encoded with a scientific objective and an exploration strategy incurs no communication delays and can fulfill missions more quickly. Active learning algorithms offer this capability of intelligent exploration, but the underlying model structure varies the performance of the active learning algorithm in accurately forming an understanding of the environment. In this paper, we investigate the performance differences between active learning algorithms driven by Gaussian processes or Bayesian neural networks for exploration strategies encoded on agents that are constrained in their trajectories, like planetary surface rovers. These two active learning strategies were tested in a simulation environment against science-blind strategies to predict the spatial distribution of a variable of interest along multiple datasets. The performance metrics of interest are model accuracy in root mean squared (RMS) error, training time, model convergence, total distance traveled until convergence, and total samples until convergence. Active learning strategies encoded with Gaussian processes require less computation to train, converge to an accurate model more quickly, and propose trajectories of shorter distance, except in a few complex environments in which Bayesian neural networks achieve a more accurate model in the large data regime due to their more expressive functional bases. The paper concludes with advice on when and how to implement either exploration strategy for future space missions.
Conducting Spectra-Spatial Investigations on the Big Island of Hawaii as a Lunar Surface Analogue
In order to evaluate the likeness of the Big Island as a lunar surface analogue to simulate lunar mission operations, this project is proposed to conduct spectra-spatial investigation with a visible and near-infrared (VNIR) spectrometer. The field spectral measurements are compared to the returned lunar samples and the lunar simulants. In this project, a procedure is developed to collect spectra in analogue field tests and a quantitative analysis of the material likeness of ground samples at the Big Island planetary analogue field site to each other (sample consistency) and to Apollo samples (lunar similarity) is conducted. This comparative analysis has provided valuable insights into the consistency and similarity of the analogue to the lunar surface, which will benefit the development of scientific and technological investigations intended for the lunar surface missions.
Wang, Hao, Krystal Arroyo-Flores, and Frances Zhu. "Conducting Spectra-Spatial Investigations on the Big Island of Hawaii as a Lunar Surface Analogue." ASCEND 2023. 2023. 4661.
Akemoto, A. and Zhu, F., 2023, March. Cooperative Lunar Surface Exploration using Transfer Learning with Multi-Agent Visual Teach and Repeat. In 2023 IEEE Aerospace Conference (pp. 1-9). IEEE.
From in-situ resource utilization to lunar habitation and rocket launch sites, autonomous robots will play a large role in supporting the infrastructure and development of permanent habitation on the Moon's surface. Current perception, localization, and path planning algorithms provide mobility to autonomous robots at the cost of heavy computation and data volume, expensive sensors, and communication limited by latency and volume. This paper aims to introduce an extension to traditional teach-and-repeat algorithms which use a teleoperation learning phase to inform an autonomous repetition phase. Rather than a single robot repeating a previously explored path, we explore the performance of a heterogenous robotic system sharing pathing data amongst different agents. A transfer-learning based teach-and-repeat system provides flexible and scalable mobility to a swarm of robots. Monocular image feed and basic odometry data is shared between robots for traversal. This research paves the way towards synchronous multi-agent algorithms that are lightweight and scalable for repetitive navigation tasks. The results show the performance of transfer teach-and-repeat algorithms across terrains of varied topological conditions, path characteristics (length, angle change, elevation gain, hazard level), and the characteristics of the repeat robots in a simulated environment. The goal of this work is to create a distributed system that will enable the mobility of robots designed for tasks beyond exploration such as mining and construction.
Recent advances in deep learning have bolstered our ability to forecast the evolution of dynamical systems, but common neural networks do not adhere to physical laws, critical information that could lead to sounder state predictions. This contribution addresses this concern by proposing a neural network to polynomial (NN-Poly) approximation, a method that furnishes algorithmic guarantees of adhering to physics while retaining state prediction accuracy. To achieve these goals, this article shows how to represent a trained fully connected perceptron, convolution, and recurrent neural networks of various activation functions as Taylor polynomials of arbitrary order. This solution is not only analytic in nature but also least squares optimal. The NN-Poly system identification or state prediction method is evaluated against a single-layer neural network and a polynomial trained on data generated by dynamic systems. Across our test cases, the proposed method maintains minimal root-mean-squared state error, requires few parameters to form, and enables model structure for verification and safety. Future work will incorporate safety constraints into state predictions, with this new model structure and test high-dimensional dynamical system data.
Zhu, Frances, et al. "NN-Poly: Approximating Neural Networks by Taylor Polynomials for Safer State Prediction." Frontiers in Robotics and AI (2022): 235.
Sloan, A., Ngo, K., Amendola, C., Clements, L., Takushi, E., Imai-Hong, A., and Zhu, F., “University of Hawaii's Spaceflight-Ready, Low-Cost, Open-Source, Educational Artemis CubeSat Kit” in 2022 SmallSat Conference Proceedings.
The Artemis CubeSat Kit is a spaceflight-ready, low-cost, educational 1U CubeSat kit, which acts as a foundation enabler in aerospace engineering education and commercial small satellites. The hardware kit accompanies a standalone “Spacecraft Mission Design” curriculum in the public domain, which includes a self-guided course outline, textbook, and digital lab modules. Funded by NASA’s Artemis Student Challenge Program, the Kit is developed and maintained by students attending the University of Hawaii at Manoa and supervised by the Hawaii Space Flight Laboratory (HSFL). The purpose of the Artemis CubeSat Kit and accompanying curriculum is to provide educational accessibility for university-level students and faculty interested in designing, building, and flying their own small satellite missions. This paper describes the technical design of the Artemis 1U CubeSat, the topology of the standalone curriculum, and the lessons learned by the team while developing the Artemis CubeSat Kit.
Project POKE (Providing Opportunities for Keiki in Engineering) is a unique opportunity for middle and high school keiki (children in Hawaiian) in Hawaii to gain hands-on aerospace experience in their classrooms through interaction with a 1U CubeSat kit. The focus of the program is to build a STEM community based on collaboration and project-based learning to offset learning loss during the COVID-19 pandemic. Highlights include an educator course, CubeSat kit, collaborative digital space, design challenge, and symposium. Middle and high school teachers concurrently participate in the educator course and meet with their students to teach the provided content. Project POKE builds off of the Artemis CubeSat Kit while modeling the course material to K-12 education. Students are challenged to develop a mission concept on how they would study a problem affecting their community using a CubeSat, and present their design concept to STEM professionals at the culminating symposium. The first iteration of the program was completed in the SY21-22 with 14 educators and over 100 students. Both the educator survey and student feedback results yielded positive sentiment regarding the program and several learning outcomes. Project POKE aims to create a diversified STEM community in Hawaii, while demystifying space science.
Ngo, K., Sloan, A., Amendola, C., Clements, L., Imai-Hong, A., and Zhu, F., “Project POKE (Providing Opportunities for Keiki in Engineering): Developing a STEM Community to Offset Learning Loss Amidst COVID Pandemic through Aerospace Technologies Project-Based Learning in Hawaii’s K-12 Classrooms” in 2022 SmallSat Conference Proceedings.
Some environments remain largely unexplored, like unvisited planetary surfaces or regions of the ocean. Many of the destinations we are interested in have never been explored and thus, there is no a priori knowledge to leverage. Instead, a vehicle must sample upon arrival, process this data, and either send this information back to a teleoperator or autonomously decide where to go next. Teleoperation is suboptimal in that human intuition can be imprecise and cannot be mathematically guaranteed to yield an optimal result. Given a surface environment, a mobile agent will map the distribution of a scalar variable without any prior information, with a degree of confidence of model convergence, and while minimizing distance traveled. Science-blind approaches to covering a surface area include a predefined path with waypoints for a vehicle to locomote in an “open-loop” policy, like the Boustrephedon or Spiral patterns. Information theoretic approaches iteratively gather information to feed back into a model or policy, which then updates with a waypoint to visit next. Gaussian processes have been popularly incorporated into active learning exploration strategies due to their ability to incorporate information into a parameter-free model and quantify a confidence metric with every model prediction. We evaluate the performance of this informative path planning algorithm in mapping an environment on three different surfaces: parabola, Townsend, and hydration value across a lunar crater from LAMP data. The Gaussian process model was tasked to learn these relationships across these specific surfaces to investigate the efficacy of the learner and policy on a noiseless vs noisy surface, a convex vs a nonconvex surface, and a smooth well-behaved function vs an unknown real function. We quantify model variance, model root-mean-squared error, distance, and surface global minimum identification of the various methods in exploring a range of surfaces with no a priori knowledge. The results show that the information-driven methods significantly outperform naive exploration methods in minimizing model error and distance with potential of convergence.
Akemoto, A. and Zhu, F., “Informative Path Planning to Explore and Map Unknown Planetary Surfaces with Gaussian Processes,” in 2022 IEEE Aerospace Conference Proceedings.
This white paper hopes to illuminate the science motivation for exploring subsurface environments, the types of subsurface environments worth exploring, and the future of robotic technology to achieve these subsurface science objectives. Technologies include robotic manipulators, diverse robot morphologies, and machine learning techniques.
Zhu, F., French, L., Schorghofer, N., Blachowicz, A., Li, S., Schurmeier, L. and Paton, M. “Robot Technology Advancements for In-Situ Exploration of Subsurface Environments”. Bulletin of the American Astronomical Society, 2021, 53(4), p.384.
We’ve created this open source, free, online textbook to bring the love and knowledge of spacecraft mission engineering to as many people as possible. This resource is free to you because the creators were funded through the NASA Artemis program. Cost of a textbook or access to a formal aerospace engineering program should not be an obstacle to your pursuit of building spacecraft. Let’s get rid of the silly notion that you need to be a “rocket scientist” to work stuff that goes to space. We’re seeing the educational barrier to building satellites drop lower and lower; middle schoolers and high schoolers have sent satellites to space [NASA]. By including as many people as possible into our community, we are fostering the most diverse and creative ideas. Inclusion pushes forward our community’s boundary of knowledge, whether that community is in your classroom or club, in your state, in your nation, or in your world. We hope that you find other soon-to-be spacecraft engineers and use this textbook to craft your own spacecraft.
Zhu, Frances, D. Sawyer Elliott, ZhiDi Yang, and Haoyuan Zheng.. "Learned and Controlled Autonomous Robotic Exploration in an Extreme, Unknown Environment." 2019 IEEE Aerospace Conference. IEEE, 2019.
Exploring and traversing extreme terrain with surface robots is difficult, but highly desirable for many applications, including exploration of planetary surfaces, search and rescue, among others. For these applications, to ensure the robot can predictably locomote, the interaction between the terrain and vehicle, terramechanics, must be incorporated into the model of the robot's locomotion. Modeling terramechanic effects is difficult and may be impossible in situations where the terrain is not known a priori. For these reasons, learning a terramechanics model online is desirable to increase the predictability of the robot's motion. A problem with previous implementations of learning algorithms is that the terramechanics model and corresponding generated control policies are not easily interpretable or extensible. If the models were of interpretable form, designers could use the learned models to inform vehicle and/or control design changes to refine the robot architecture for future applications. This paper explores a new method for learning a terramechanics model and a control policy using a model-based genetic algorithm. The proposed method yields an interpretable model, which can be analyzed using preexisting analysis methods. The paper provides simulation results that show for a practical application, the genetic algorithm performance is approximately equal to the performance of a state-of-the-art neural network approach, which does not provide an easily interpretable model.
For pre-print link outside of any paywalls, click here.
Zhu, Frances, and Mason A. Peck. "Linearized dynamics of general flux-pinned interfaces." IEEE Transactions on Applied Superconductivity 28.8 (2018): 1-10.
A flux-pinned interface offers a passively stable equilibrium that otherwise cannot occur between magnets because electromagnetic fields are divergenceless. The contactless, compliant nature of flux pinning offers many benefits for close-proximity robotic maneuvers, such as rendezvous, docking, and actuation. This paper derives the six degree-of-freedom linear dynamics about an equilibrium for any magnet/superconductor configuration. Linearized dynamics are well suited to predicting close-proximity maneuvers, provide insights into the character of the dynamic system, and are essential for linear control synthesis. The equilibria and stability of a flux-pinned interface are found using Villani’s equations for magnetic dipoles. Kordyuk’s frozen-image model provides the nonlinear flux-pinning response to these magnetic forces and torques, all of which are then linearized. Comparing simulation results of the nonlinear and linear dynamics shows the extent of the linear model's applicability. Nevertheless, these simple models offer computational speed and physical intuition that a nonlinear model does not.
For pre-print link outside of any paywalls, click here.
Zhu, Frances, Mason Peck, and Laura Jones-Wilson. "Flux-pinned dynamics model parameterization and sensitivity study." Journal of Aerospace Information Systems (2019): 1-16.
Flux-pinned interfaces maintain a passively stable equilibrium between two spacecraft in close proximity. Although flux-pinning physics has been studied from a materials science perspective and at the system level, the sensitivities and implications of system-level designs on the dynamics need to be better understood, especially in interfaces with multiple magnets and superconductors. These interfaces have highly nonlinear coupled dynamics that are influenced by physical parameters, including strength of magnetic-field sources, field-cooled position, and superconductor geometry. Kordyuk’s frozen-image model (“Magnetic Levitation for Hard Superconductors,” Journal of Applied Physics, Vol. 83, No. 1, Jan. 1998, pp. 610–612) successfully approximates the characteristics of flux-pinning dynamics, but could provide a more precise state prediction with the addition of these physical parameter refinements. This paper addresses that gap by offering parametric terms to improve the dynamics model, which may better simulate the behavior of a multiple-magnet-and-multiple-superconductor interface. The sensitivity of the general flux-pinned dynamics model is studied by varying the physical parameters and simulating the system-level dynamics. This work represents a critical step in the development of a model suited to spacecraft performance verification.
For pre-print link outside of any paywalls, click here.
Zhu, Frances, Laura Jones-Wilson, and M. Peck. "A concept for capturing and docking spacecraft with flux-pinned interfaces." 67th International Astronautical Congress. 2016.
This paper describes a set of air-bearing experiments and simulations designed to characterize the dynamics of an Orbiting Sample (OS) capture by a Flux-Pinned Interface (FPI) and determine the bounds over which the interface will generate a successful capture. The experiments were performed with an OS analogue on a 4 degree-of-freedom planar air bearing carriage. This unit was flux pinned to a stationary sample return orbiter analogue, and data was collected over a variety of different capture scenarios, especially over variations in the magnitude of incoming velocity. The simulation models the physics of both the testbed and an on-orbit system and is used to extend these empirical results to a flight FPI. It was determined that slower on-axis velocities (3.5 cm/s) nearly guaranteed capture, but velocities above 6 cm/s did not consistently capture. The experiments also show a sensitivity to variation in the number of degrees of freedom, which suggests a microgravity flight may be a better testing environment. The simulation shows that the existing models of flux pinning physics over-predict the performance of the capture system and need to be modified to take into account various system limitations. The paper concludes with the implications of this research for sample capture applications and paths for future technology development.
For pre-print link outside of any paywalls, click here.
Zhu, Frances, Mitchell Dominguez, Mason Peck, and Laura Jones-Wilson. "Flight-experiment validation of the dynamic capabilities of a flux-pinned interface as a docking mechanism." 2019 IEEE Aerospace Conference. IEEE, 2019.
Flux-pinned interfaces for spacecraft leverage the physics of superconductor interactions with electromagnetism to govern the dynamics between two bodies in close-proximity. Several unique advantages over traditional mechanical capture systems include robustness to control failures, contactless reorientation of the capture target, and collision mitigation. This study describes a series of experiments performed in a microgravity environment during a parabolic-flight campaign to measure the dynamic behavior of a flux-pinned interface in a flight-traceable environment. This paper presents the performance of a flux-pinned interface in the full six degrees of freedom in terms of several quantifiable metrics: success of capture at various energetic states, momentum change, system damping, and interface stiffness of the two spacecraft bodies.
For pre-print link outside of any paywalls, click here.
Zhu, Frances, Mason Peck, and Laura Jones-Wilson. “Reduced Embedded Magnetic Field in Type-II Superconductor of Finite Dimension.” IEEE Transactions on Applied Superconductivity, 2020, 1–1. https://doi.org/10.1109/TASC.2020.2976592.
This work maps the magnetic field within a type-II superconductor of finite dimension that is magnetically flux-pinned. The measured field is lower in magnitude than anticipated from the frozen image model and changes shape dependent on the field-cooled image location. A proposed refined model more accurately reflects the measured field.
For pre-print link outside of any paywalls, click here.