publications

------------------------------------------

Journal Papers

1. TRO2024

   MAGIC-VFM: Meta-learning Adaptation for Ground Interaction Control with Visual Foundation Models

   E. S. Lupu*, F. Xie*, J. A. Preiss, J. Alindogan, M. Anderson, and S.-J. Chung

   IEEE Transactions on Robotics, 2024,

   Abs Video

   Control of off-road vehicles is challenging due to the complex dynamic interactions with the terrain. Accurate modeling of these interactions is important to optimize driving performance, but the relevant physical phenomena are too complex to model from first principles. Therefore, we present an offline meta-learning algorithm to construct a rapidly-tunable model of residual dynamics and disturbances. Our model processes terrain images into features using a visual foundation model (VFM), then maps these features and the vehicle state to an estimate of the current actuation matrix using a deep neural network (DNN). We then combine this model with composite adaptive control to modify the last layer of the DNN in real time, accounting for the remaining terrain interactions not captured during offline training. We provide mathematical guarantees of stability and robustness for our controller, and demonstrate the effectiveness of our method through simulations and hardware experiments with a tracked vehicle and a car-like robot. We evaluate our method outdoors on different slopes with varying slippage and actuator degradation disturbances, and compare against an adaptive controller that does not use the VFM terrain features. We show significant improvement over the baseline in both hardware experimentation and simulation.

1. SciRob2021

   A bipedal walking robot that can fly, slackline, and skateboard

   K. Kim, P. Spieler, E. S. Lupu, A. Ramezani, and S.-J. Chung

   Science Robotics, 2021,

   Abs Video

   Numerous mobile robots in various forms specialize in either ground or aerial locomotion, whereas very few robots can perform complex locomotion tasks beyond simple walking and flying. We present the design and control of a multimodal locomotion robotic platform called LEONARDO, which bridges the gap between two different locomotion regimes of flying and walking using synchronized control of distributed electric thrusters and a pair of multijoint legs. By combining two distinct locomotion mechanisms, LEONARDO achieves complex maneuvers that require delicate balancing, such as walking on a slackline and skateboarding, which are challenging for existing bipedal robots. LEONARDO also demonstrates agile walking motions, interlaced with flying maneuvers to overcome obstacles using synchronized control of propellers and leg joints. The mechanical design and synchronized control strategy achieve a unique multimodal locomotion capability that could potentially enable robotic missions and operations that would be difficult for single-modal locomotion robots.

2. ASR2021

   Decentralized formation pose estimation for spacecraft swarms

   K. Matsuka, A. O. Feldman, E. S. Lupu, S.-J. Chung, and F. Y. Hadaegh

   Advances in Space Research, 2021,

   Abs Video

   For spacecraft swarms, the multi-agent localization algorithm must scale well with the number of spacecraft and adapt to time-varying communication and relative sensing networks. In this paper, we present a decentralized, scalable algorithm for swarm localization, called the Decentralized Pose Estimation (DPE) algorithm. The DPE considers both communication and relative sensing graphs and defines an observable local formation. Each spacecraft jointly localizes its local subset of spacecraft using direct and communicated measurements. Since the algorithm is local, the algorithm complexity does not grow with the number of spacecraft in the swarm. As part of the DPE, we present the Swarm Reference Frame Estimation (SRFE) algorithm, a distributed consensus algorithm to co-estimate a common Local-Vertical, Local-Horizontal (LVLH) frame. The DPE combined with the SRFE provides a scalable, fully-decentralized navigation solution that can be used for swarm control and motion planning. Numerical simulations and experiments using Caltech’s robotic spacecraft simulators are presented to validate the effectiveness and scalability of the DPE algorithm.

1. ActaAstro2020

   Autonomous in-orbit satellite assembly from a modular heterogeneous swarm

   R. C. Foust, E. S. Lupu, Y. K. Nakka, S.-J. Chung, and F. Y. Hadaegh

   Acta Astronautica, 2020,

   Abs Video

   This paper presents a decentralized, distributed guidance and control scheme to combine a heterogeneous swarm of component satellites into a large satellite structure. The component satellites for the heterogeneous swarm are chosen to promote flexibility in final shape inspired by crystal structures and Islamic tile art. After the ideal fundamental building blocks are selected, basic nanosatellite-class satellite designs are made to assist in simulations involving attitude control. The Swarm Orbital Construction Algorithm (SOCA) is a guidance and control algorithm to allow for the limited type heterogeneity and docking ability required for in-orbit assembly. The algorithm consists of two parts, a distributed auction which uses barrier functions to ensure the proper agent selection for each target, and a trajectory generation portion which leverages model predictive control and sequential convex programming to achieve optimal collision-free trajectories to the desired target point even with nonlinear system dynamics. The optimization constraints use a boundary layer to determine whether the collision avoidance or the docking constraints should be applied. The algorithm was tested in a simulated perturbed 6-DOF spacecraft dynamic environment for planar and out-of-plane final structures and on two robotic platforms, including a swarm of frictionless spacecraft simulation robots.

1. IOP2015

   The GBT-FPGA core: features and challenges

   M. Barros Marin, S. Baron, S.S. Feger, P. Leitao, E. S. Lupu, C. Soos, and 2 more authors

   Journal of Instrumentation, 2015,

   Abs

   Initiated in 2009 to emulate the GBTX (Gigabit Transceiver) serial link and test the first GBTX prototypes, the GBT-FPGA project is now a full library, targeting FPGAs (Field Programmable Gate Array) from Altera and Xilinx, allowing the implementation of one or several GBT links of two different types: "Standard" or "Latency-Optimized". The first major version of this IP Core was released in April 2014. This paper presents the various flavours of the GBT-FPGA kit and focuses on the challenge of providing a fixed and deterministic latency system both for clock and data recovery for all FPGA families.

------------------------------------------

Conference Papers

1. SmallSats2024

   Vision-Based Detection of Uncooperative Targets and Components on Small Satellites

   H. Grauer*, E. S. Lupu*, C. Lee, D. Rowen, B. Bycroft, P. Leeds, and 2 more authors

   38th Annual Small Satellite Conference, 2024,

   Abs

   Space debris and inactive satellites pose a threat to the safety and integrity of operational spacecraft and motivate the need for space situational awareness techniques. These uncooperative targets create a challenging tracking and detection problem due to a lack of prior knowledge of their features, trajectories, or even existence. Recent advancements in computer vision models can be used to improve upon existing methods for tracking such uncooperative targets to make them more robust and reliable to the wide-ranging nature of the target. This paper introduces an autonomous detection model designed to identify and monitor these objects using learning and computer vision. The autonomous detection method aims to identify and accurately track the uncooperative targets in varied circumstances, including different camera spectral sensitivities, lighting, and backgrounds. Our method adapts to the relative distance between the observing spacecraft and the target, and different detection strategies are adjusted based on distance. At larger distances, we utilize You Only Look Once (YOLOv8), a multitask Convolutional Neural Network (CNN), for zero-shot and domain-specific single-shot real time detection of the target. At shorter distances, we use knowledge distillation to combine visual foundation models with a lightweight fast segmentation CNN (Fast-SCNN) to segment the spacecraft components with low storage requirements and fast inference times, and to enable weight updates from earth and possible onboard training. Lastly, we test our method on a custom dataset simulating the unique conditions encountered in space, as well as a publicly-available dataset.

1. ICRA2020

   Adaptive Nonlinear Control of Fixed-Wing VTOL with Airflow Vector Sensing

   X. Shi, P. Spieler, E. Tang, E. S. Lupu, P. Tokumaru, and S.-J. Chung

   IEEE International Conference on Robotics and Automation (ICRA), 2020,

   Abs Video

   Fixed-wing vertical take-off and landing (VTOL) aircraft pose a unique control challenge that stems from complex aerodynamic interactions between wings and rotors. Thus, accurate estimation of external forces is indispensable for achieving high performance flight. In this paper, we present a composite adaptive nonlinear tracking controller for a fixed-wing VTOL. The method employs online adaptation of linear force models, and generates accurate estimation for wing and rotor forces in real-time based on information from a three-dimensional airflow sensor. The controller is implemented on a custom-built fixed-wing VTOL, which shows improved velocity tracking and force prediction during the transition stage from hover to forward flight, compared to baseline flight controllers.

1. AIAA2018

   Six Degree-of-Freedom Spacecraft Dynamics Simulator for Formation Control Research

   Y. Nakka, R. Foust, E. S. Lupu, D. B. Elliott, I. S. Crowell, S.-J. Chung, and 1 more author

   AAS/AIAA Astrodynamics Specialist Conference, 2018,

   Abs Video

   This paper presents a new six-degree-of-freedom robotic spacecraft simulator, the Multi-Spacecraft Testbed for Autonomy Research (M-STAR), for testing formation guidance, relative navigation, and control algorithms. The simulator dynamics are governed by five degrees of frictionless translational and rotational air-bearing motion and one degree of kinematic motion in the gravity direction with flightlike actuators, in a 1-g environment. The simulator is modelled as a 3-D pendulum on a floating platform with sixteen thrusters and four reaction wheels as on-board actuators. Based on this plant model, a nonlinear hierarchical control law is proposed for position and attitude trajectory tracking. A weighted generalized pseudo-inverse strategy for control allocation to map control inputs to actuator inputs is discussed. The thruster actuation model for mapping smooth allocated input to non-smooth actuator input that achieves equivalent performance is derived. The control law and allocation scheme are tested on the simulator for real-time position control using the proposed actuator model.

2. IAC2018

   Ultra-soft electromagnetic docking with applications to in-orbit assembly

   R. Foust, E. S. Lupu, Y. Nakka, D. B. Elliott, I. S. Crowell, S.-J. Chung, and 1 more author

   International Astronautical Congress, 2018,

   Abs Video

   Docking small satellites in space is a high-risk operation due to the uncertainty in relative position and orientation and the lack of mature docking technologies. This is particularly true for missions that involve multiple docking and undocking procedures like swarm-based construction and reconfiguration. In this paper, an electromagnetic docking system is proposed to mitigate these risks through robust, ultra-soft, propellant-free docking. Designed with reconfigurable self-assembly in mind, the gripping mechanism is androgynous, able to dock at a variety of relative orientations, and tolerant of small misalignments. The mechanical and control design of the system is presented and tested in both simulation and on a fleet of 6 degree-of-freedom (DOF) spacecraft simulators. The spacecraft simulators float on the precision flat floor facility in the Caltech Aerospace Robotics and Control lab, the largest of its kind at any university. The performance of the electromagnetic docking system on-board the simulators is then compared against a propulsive docking system.

1. IAC2013

   Investigation of the Surface Deformation and Dendritic Solidification of a Titanium Allow melted in Miligravity

   E. S. Lupu, I. Ciuca, L. Manoliu, C. Chitu, C. Cherciu, I. Ciobanu, and 4 more authors

   Hans Von Muldau Award for the Best Team Project

   International Astronautical Congress (IAC), 2013,

   Abs Video

   With the construction of the International Space Station, microgravity research has enabled many scientific breakthroughs. Moreover, new materials can be synthesized under low gravity conditions. In particular, in our analysis, an investigation on the surface deformation and on the dendritic solidification of a titanium alloy melted under Earth-based laboratory conditions and on board of a sounding rocket is performed. This paper represents the first part of an ample research and is concerned with describing the rocket module platform that will fly during the REXUS 15/16 campaign from Kiruna, Sweden and the results obtained during the on-ground experiment. Also, it includes a detailed research regarding the hypothetical differences that will be obtained during the sub-orbital flight. The experiment contains a multimode LASER diode with an output power of 25 W that will melt samples Ti6Al4V covered with nanotubes. The nanotubular array is elaborated in Polyethylene glycol (PEG600). The research follows to understand the changes in microstructure, especially in the dendritic distribution. The results obtained under Earth conditions are presented in the Results section and then commented on the Discussion section. Here, we propose a comparative analysis with the results obtained under reduced gravity conditions (in the rocket), where we expect a difference in the dendrite geometry, particularly an increased spacing in the dendrites of an alloy that solidify in low gravity. The paper also contains a mathematical model of the heat transfer mechanism that is purely diffusive. The second part of our research is concerned with the experiment on board of the rocket, when, during the 120 seconds of miligravity, the LASER is expected to melt the samples. The environmental parameters are going to be monitored using sensors and the melting process is recorded with a camera. In conclusion, we take into consideration the fact that the only parameter that differs between the on-ground and the on board of the rocket experiment shall be the level of gravity (the pressure, the temperature and the humidity must remain the same). So, a proficient comparison between the low and normal gravity effects can be undertaken.

------------------------------------------

Workshop Papers

1. RSS2025

   Vision-Based Detection of Uncooperative Targets and Components on Small Satellites

   H. Grauer*, E. S. Lupu*, C. Lee, D. Rowen, B. Bycroft, P. Leeds, and 2 more authors

   Best Paper Award

   Space Robotics Workshop at the Robotics: Science and Systems Conference,
2. RSS2025

   MAGIC-VFM: Meta-learning Adaptation for Ground Interaction Control with Visual Foundation Models

   E. S. Lupu*, F. Xie*, J. A. Preiss, J. Alindogan, M. Anderson, and S.-J. Chung

   Best Poster Runner-Up Award

   Resilient Off-road Autonomous Robotics (ROAR) Workshop at the Robotics: Science and Systems Conference,
3. ICRA2025

   MAGIC-VFM: Meta-learning Adaptation for Ground Interaction Control with Visual Foundation Models

   E. S. Lupu*, F. Xie*, J. A. Preiss, J. Alindogan, M. Anderson, and S.-J. Chung

   Second Place

   Workshop on Foundation Models and Neuro-Symbolic AI for Robotics at IEEE International Conference on Robotics and Automation (ICRA), 2025,

   Video

1. IWSCFF2019

   Distributed multi-target relative pose estimation for cooperative spacecraft swarm

   K. Matsuka, E. S. Lupu, Y. Nakka, R. Foust, S.-J. Chung, and F. Y. Hadaegh

   Best IWSCFF Student Paper Award

   10th International Workshop on Satellite Constellations and Formation Flying,

   Abs Video

   Multi-agent relative state estimation is critical in enabling full swarm autonomy. However, relative pose estimation of hundreds to thousands of cooperative agents is challenging due to limited sensing, limited communication, and scalability. We present a distributed algorithm for cooperative multi-agent localization with both limited relative sensing and communication. Each agent locally exchanges the relative measurements and jointly estimates the relative poses of its local neighbors. Because the algorithm only estimates the local neighbors, the number of states does not grow with the total number of agents given the same local sensing and communication graphs, making the algorithm suitable for swarm application. The proposed algorithm is applied to spacecraft swarm localization and verified in simulation and experiments. Experiments are conducted on Caltech’s robotic spacecraft simulators, the Multi-Spacecraft Testbed for Autonomy Research (M-STAR), where each spacecraft uses vision-based relative measurements.