Free Flyer Guidance

The problem

An intravehicular free flyer must hold and change a six degree of freedom state inside a volume it can collide with, without GPS, without a fixed base, and on a processor small enough to fly inside a pressurized module. The three families that have operated on the ISS solve the navigation half differently: ultrasonic ranging, fiducial markers, and visual localization against a prebuilt map [2][1]. The guidance problem changed with each step from thruster-limited cold gas vehicles on a beacon grid to fan-driven vehicles whose limits are acoustic noise, crew safety and compute [16][7].

None of the guidance can be tested at full six degrees of freedom on Earth. Astrobee uses a Gazebo simulator plus a granite table facility with two ground models, CIMON used software simulation, ground model testing, an elegant breadboard and hardware-in-the-loop simulation, and Int-Ball2 uses Gazebo, a granite table with a ground model, and hardware-in-the-loop simulation [1]. The granite table constrains motion to a plane, which is why SPHERES-era results distinguish between 2D table results and on-orbit six degree of freedom results [2].

What has flown

SPHERES and its guest-scientist work

SPHERES localized against fixed ultrasonic beacons, which is why its operating volume was a 2 m cube [3]. Global metrology ran at a maximum 5 Hz update rate, giving position and 1 to 2 degrees of orientation, which caps estimator bandwidth at 2.5 Hz [2]. The control loop ran at 50 Hz and the navigation modules sub-sampled it. Each satellite masses 4.3 kg and each thruster produces 0.09 N [2]. Bench measurement on the flight design gave 0.1783 N per thruster at the 55 psig regulator maximum against 0.2480 N predicted, and per-thruster thrust fell by over 30 percent with six thrusters open, so quoted values depend on the regulator setting and on how many thrusters fire together [23].

Static estimator performance was measured with the satellite held still: all position estimates fell inside a 2 mm by 2 mm box, and experiments showed that state errors larger than 10 cm degrade behavior, that the glideslope algorithm needs an initial error under 15 cm on the axes it controls, and that with errors above 20 cm in position or 45 degrees in attitude the approach does not work [2]. The flight result was the first on-orbit autonomous docking to a tumbling target, over 15 successful experiments in the fifth and sixth ISS test sessions, at a target tumble rate of 2.25 deg/s from 40 cm separation, with terminal range within 3 cm and tangential error within 2 cm [2].

Guest-scientist extensions moved the testbed toward servicing and assembly. The 2000 design thesis reported about 10 minutes of continuous station keeping per propellant tank and a control rate cut from 50 Hz to 10 Hz after testing [10]. The estimator and control interface work developed pulse modulation for the on/off thrusters and a Kalman filter for position and velocity; the author notes the process noise was not well characterized [13]. A stereo payload supported an offline dynamic factor-graph SLAM of a spinning target from ISS Test Session 38, agreeing with ultrasound metrology to 1.59 cm mean on the range axis [12]. A campaign on electromagnetic formation flight is documented as largely failing to produce the intended data, because of a firmware fault found late and experiments that combined effects without the downlink needed to separate them [22]. A reconfigurable estimator for a tug that changes mass properties reached plus or minus 1 cm in flat-table tests, pooling the receivers of two docked satellites made the estimate worse in hardware, and the on-orbit assembly test was inconclusive [21]. A docking-port optical sensor needed better than 1 cm and 2 degrees before contact, and its results come from a flat floor and reduced-gravity aircraft, not the station [20].

Smart SPHERES attached a smartphone to a SPHERES satellite and was commanded from the ground to survey the Kibo forward wall on 12 December 2012 [11]. Ground control switched between supervisory and manual teleoperation across two loss-of-signal periods, but only one of five survey plans ran start to finish nominally, two ended on propellant exhaustion, path overshoot missed about 1 percent of the surveyed area, and the smartphone’s mass asymmetry made the camera lag and then swing at each departure [11]. These results are the operational argument the pre-launch Astrobee statement of purpose makes against SPHERES: ultrasonic beacons confined it to a 2 m cube, and tank and battery swaps were manual [7].

Astrobee

Astrobee has no ranging infrastructure. A sparse feature map of the module interior is built offline from imagery, and the robot localizes against it in flight [3]. SURF features are used for map building because they are expensive but computed offline, while BRISK features are used onboard. The original estimator was an augmented-state indirect extended Kalman filter fusing map feature matches, AR tags near the dock, handrail measurements from a depth sensor when perching, optical flow for velocity, and the IMU. WiFi was evaluated as a localization modality and rejected because the existing ISS access point geometry does not give sufficient position accuracy. Granite table testing gave position RMSE between about 4.9 and 18 cm and angular RMSE between about 1.3 and 14 degrees depending on configuration [3].

The operational problem was drift and lost events when the prior map does not match a changed environment, and AstroLoc replaced the MSCKF-based filter with graph-based optimization on Astrobee’s InForce 6501 Micro SoM, a single core that runs roughly 10 times slower than an Intel i9-9980HK at 2.4 GHz and shares the machine with other processes [4]. Reduced reliance on the prior map cut the activities run specifically to collect new map imagery by about 40 percent and the images per such activity by about 30 percent, and Astrobee has since executed ISS activities with over two hours of flying time [4], against a 3 hour battery endurance [1]. The primary record counts over 165 real-time operations and over 90 Guest Science test sessions through May 2025 [5].

Int-Ball, Int-Ball2, CIMON

Fan propulsion changed the guidance problem from thruster allocation to low-thrust control against ISS airflow. First-generation Int-Ball fans produced roughly 1 mN each, which was insufficient to hold pose in regions of heavy ventilation, and Int-Ball2 was redesigned with eight large propeller modules and about 0.15 N on its dominant axis [16]. The thrust figures in that survey are defined differently across platforms, so they do not rank vehicles directly [16].

Int-Ball2 flew its first on-orbit demonstration on 17 October 2023 against requirements of 95 mm and 11 degrees for navigation and 30 mm and 2.5 degrees for control [17]. The reported control error on a +X translation was 1.65 mm in orbit against 8.96 mm on the ground hardware-in-the-loop simulator, while a rotation produced a navigation error of 91.64 mm; the orbital ground truth came from reference markers with their own error of up to 28 mm and 1.1 degrees, and the orbital sample is one session with two docking attempts [17].

CIMON, built by Airbus for DLR, launched as CIMON-2 on CRS-19 on 5 December 2019 [24] and, in first tests with an ESA astronaut, flew to a named location in Columbus from an arbitrary start using absolute navigation. The source is a manufacturer press release with no position accuracy, so CIMON guidance performance is not quantified in the held literature.

Flown only on the ground

The Mini AERCam, a 7.5 in, 10 lb sphere with twelve cold gas thrusters, a 266 MHz PowerPC 740 avionics board and a Kalman filter publishing position and velocity at 1 Hz, was built to succeed the 1997 AERCam Sprint, which flew on STS-87 [15]. Mini AERCam itself was tested on an air-bearing table (two translational axes and one rotational axis) and in orbital simulation, not in orbit [15]. Its software separated a CLIPS expert-system command sequencer, a system manager for anomaly response and a dynamics manager that commands contingency maneuvers, again with no flight data [14].

Failure modes in operation

Losing the position fix is the dominant one, and its consequence is crew intervention [4]. On Astrobee the localization fix was occasionally lost during activities, initially mitigated by returning the robot to its dock and renewing the fix by visiting mapped areas, and later by updating the localization algorithms to tolerate environmental change [1][4]. Int-Ball2’s relative pose estimation accumulates error, and post-flight analysis shows about 8 percent of flight time encountering errors that produce a localization shift; the recommendation is absolute navigation references [1]. The AstroLoc navigation state is checked against a covariance threshold and a pose history check against map-derived poses, and the localizer resets if either fails [4].

Infrastructure fails alongside the robot. Separate Astrobee dock anomalies involving its Ethernet connection, plus on-orbit Wi-Fi behavior, caused loss of communication during operations; Int-Ball2 added a wired UART option while docked in response [1].

What is being tried now

Localization that tolerates a changing module

All of these run on the Astrobee flight computer on the ISS. AstroLoc is flown and measured (see the tables below) [4][5]. AstroLoc2 decouples odometry from absolute localization into a loosely coupled sequential pipeline, so noise in the rearranged ISS map no longer degrades velocity and IMU bias estimates, and adds time-of-flight correspondences and adaptive map matching. On a public dataset of 10 ISS activities it reduced absolute trajectory error RMSE by 16 percent and raised success rate by 5.5 percent over AstroLoc [25], of which sequential localization alone accounts for 4.3 percent of the RMSE reduction. It is less accurate than DSO and ORB-SLAM3 but runs 41 and 87 percent faster [25]. Semantic localization uses a learned detector of 8 object classes in the JEM as additional factors in the pose graph; it is described as the first use of learned semantics for absolute localization of a free flyer, and on five ISS datasets adding semantics to sparse features lowered position RMSE in every one [26]. Both were evaluated on recorded ISS data.

Learned control

Two reports describe reinforcement-learning controllers flown on Astrobee. A PPO policy trained in Isaac Lab and checked in Gazebo and on the Granite Lab air bearing was run on the Bumble Astrobee on 27 May 2025 in a roughly 7 minute ISS window of 8 maneuvers [9]. The policy had larger final position and orientation error than the baseline controller in every reported test, the first of two docking attempts failed with a localization problem suspected but not established, and no safety guarantee is claimed [9]. A second pipeline, trained in Omniverse with mass curriculum learning, flew one policy; the only flight maneuver shown is a single 0.5 m undock translation that undershot the target, and success statistics come from the training simulator rather than hardware [29]. Both are proofs of concept, not replacements for the classical controller. In simulation only, a symmetry-reduced PPO tracking controller for an Astrobee model reached a mean position error of 1.3 cm with a standard deviation of 2.7 cm, with no sensor noise or actuator model, no hardware and no granite-table test [27]. Docking by RL has also been studied in simulation for the Int-Ball2 plant [6].

Inspection autonomy and servicing

An autonomy architecture for a small cold-gas inspector near an orbital outpost combines operator-specified inspection points, keep-in corridors, keep-out convex hulls and a receding-horizon planner. In simulation against approximate Lunar Gateway geometry, linger mode cost roughly five times the translational impulse of flyby over the same 41.32 m path [28], and four simultaneous thruster failures drove the flyer into the station. There is no hardware, no orbital dynamics and perfect localization is assumed [28].

Multi-robot, survey and manipulation applications

A two-robot change-detection scheme, with one Astrobee building a 3D model and another scanning for inconsistencies, ran on surveys in the Granite Lab and on six ISS surveys, but no detection rate or position error is reported, and the authors expect roughly a ten-fold slowdown on the flight processor relative to a desktop [18]. Astrobee Wi-Fi survey flights in 2022 found that a free-space inverse-square model does not hold inside the modules: the fitted pathloss exponent along the longitudinal axis of the JEM was 0.084, and the radio was not characterized before flight, so only relative signal levels are reliable [19]. Mass-limited manipulation is a guidance constraint: in simulation, an Astrobee-class flyer with a 2 DOF arm held end-effector error near 0.95 mm only with base stabilization torque above the assumed 0.02 N m limit for heavier cargo bags [16].

Software as the product

Astrobee has four payload bays, of which three can carry guest hardware at once, each supplying 14.4 VDC battery power and USB 2.0 data through a blind-mate connector [1][5]. Guest investigations have spanned autonomous rendezvous and docking between two Astrobees, mapping and change detection, acoustic mapping, orbital debris capture, path planning, novel grippers, and propellant-free motion in which the perching arm tosses the robot between handrails [5]. Int-Ball2 exposes a technology demonstration platform in which user software is uplinked and started under Docker from the ground [1].

The numbers

	Astrobee	CIMON-1 and 2	Int-Ball	Int-Ball2
Developer	NASA	DLR	JAXA	JAXA
Deployed	2019 to present	2018; 2019 to present	2017 to 2019	2023 to present
Size and shape	30 cm cube	32 cm sphere	15 cm sphere	20 cm sphere
Mass	about 10 kg	about 5 kg	about 1 kg	about 3 kg
Actuation	12 nozzles fed by centrifugal impellers, 300 mN maximum x-nozzle	12 propelling fans, about 120 mN maximum in x	12 fans at about 3 mN each plus 3 reaction wheels	8 propelling fans, 60 to 163 mN each
Navigation	graph localizer	stereo camera VSLAM plus IMU fusion	monocular camera, marker based	stereo visual-inertial SLAM, feature based plus IMU fusion
Compute architecture	tiered onboard, LLP, MLP and HLP	two onboard processors plus a ground AI server	single onboard processor	single onboard processor
Processors	Wandboard Dual, InForce IFC6501, InForce IFC6601	Pokini F WIFI, AMD A4-6700T	Armadillo-810	Jetson TX2
Software	Ubuntu and ROS on LLP and MLP, Android on HLP	Ubuntu and ROS	Armadillo Base OS	Ubuntu and ROS
Power	Li-ion, automatic docking and recharge	Li-ion, crew-replaceable or wired	Li-ion, USB charge by crew	Li-ion, automatic docking and recharge
Operating time on batteries	3 h	2 h and 3 h	2 h	3 h

Source: [1], Table 1.

AstroLoc design parameters, set by the single-core budget:

Parameter	Value
Camera rate	15 Hz
IMU rate	62.5 Hz
Sliding window graph duration	3.5 s
Maximum smart factors in the factor graph	12
Maximum measurements per smart factor	5
Localizer output rate	about 5 Hz, extrapolated to 62.5 Hz by the IMU augmentor
Huber loss threshold	1.345

Measured over 12 ISS activities [4], with a lost event defined as drift exceeding 1.5 m:

Metric	Previous localizer	Base GTSAM	AstroLoc
Position RMSE, localization mode	0.0948 m	0.2417 m	0.0491 m
Orientation RMSE, localization mode	0.0439 rad	0.0505 rad	0.0275 rad
Position RMSE, VIO mode	0.6212 m	0.5328 m	0.2777 m
Position RMSE with outliers, VIO mode	2.4179 m	0.7508 m	0.3721 m
Lost events, VIO mode	2	2	0
Average runtime, VIO	not applicable, fixed 62.5 Hz	1.0588 s	0.1785 s
Average optimization time, VIO	not applicable	0.8525 s	0.1259 s
Average runtime, VIO plus map	not applicable	1.1335 s	0.2050 s

Source for both AstroLoc tables: [4]. AstroLoc runs about six times faster than the base GTSAM implementation, an 83 percent runtime decrease and an 85 percent optimization time decrease in VIO mode. SPHERES’ 5 Hz metrology capped estimator bandwidth at 2.5 Hz, while Astrobee’s constraint is the optimization time available per 15 Hz image on one core [2][4].

What is not established

Astrobee on-orbit station keeping and pointing accuracy are described only qualitatively in the primary record, and the AstroLoc errors are measured against a ground truth built by the authors’ own mapping pipeline, so the absolute error floor is not independent, with 12 activities and no per-activity spread [5]. Pre-launch Astrobee design papers contain no measurements [7][8]. The Int-Ball2 orbital result is one session [17]. The learned controllers trailed the baseline or rest on a single flight maneuver, and have no verification argument [9][29]. CIMON, Mini AERCam, the SPHERES docking-port sensor and the reconfigurable-tug estimator lack either flight data or any published position accuracy [24][15][20][21]. The inspection architecture and the symmetry-based controller exist only in simulation [28][27]. No held source compares all the flyers’ guidance on a common metric, and the thrust figures in the survey literature are not defined consistently [16].

References

Yamaguchi, S. P., Mora Vargas, A., Eisenberg, T., Rogon, C., Yamamoto, T., Inoue, S., Kössl, C., Coltin, B., Smith, T. and Benavides, J. V. (2025). Free-Flying Crew Cooperative Robots on the ISS: A Joint Review of Astrobee, CIMON, and Int-Ball Operations . International Conference on Space Robotics. Source

BibTeX

@inproceedings{yamaguchi2026free,
  title = {Free-Flying Crew Cooperative Robots on the ISS: A Joint Review of Astrobee, CIMON, and Int-Ball Operations},
  author = {Yamaguchi, Seiko Piotr and Mora Vargas, Andres and Eisenberg, Till and Rogon, Christian and Yamamoto, Tatsuya and Inoue, Shona and Kössl, Christoph and Coltin, Brian and Smith, Trey and Benavides, Jose V.},
  booktitle = {International Conference on Space Robotics},
  pages = {402-409},
  year = {2025},
  doi = {10.1109/isparo66239.2025.11436992},
  abstract = {Intra-vehicular free-flying robots are anticipated to support various work in human spaceflight while working side-by-side with astronauts. Such example of robots includes NASA’s Astrobee, DLR’s CIMON, and JAXA’s Int-Ball, which are deployed on the International Space Station. This paper presents the first joint analyses of these robot’s shared experiences, co-authored by their development and operation team members. Despite the different origins and design philosophies, the development and operations of these platforms encountered various convergences. Hence, this paper presents a detailed overview of these robots, presenting their objectives, design, and onboard operations. Hence, joint lessons learned across the lifecycle are presented, from design to on-orbit operations. These lessons learned are anticipated to serve for future development and research as design recommendations.}
}

Nolet, S. (2007). Development of a Guidance, Navigation and Control Architecture and Validation Process Enabling Autonomous Docking to a Tumbling Satellite. Source

BibTeX

@phdthesis{nolet2007development,
  title = {Development of a Guidance, Navigation and Control Architecture and Validation Process Enabling Autonomous Docking to a Tumbling Satellite},
  author = {Nolet, Simon},
  school = {Massachusetts Institute of Technology},
  type = {Ph.D. thesis},
  year = {2007},
  url = {https://dspace.mit.edu/handle/1721.1/38598}
}

Coltin, B., Fusco, J., Moratto, Z., Alexandrov, O. and Nakamura, R. (2016). Localization from Visual Landmarks on a Free-Flying Robot . IEEE/RSJ International Conference on Intelligent Robots and Systems. Source

BibTeX

@inproceedings{coltin2016localization,
  title = {Localization from Visual Landmarks on a Free-Flying Robot},
  author = {Coltin, Brian and Fusco, Jesse and Moratto, Zack and Alexandrov, Oleg and Nakamura, Robert},
  booktitle = {IEEE/RSJ International Conference on Intelligent Robots and Systems},
  pages = {4377-4382},
  address = {Daejeon, Korea},
  year = {2016},
  doi = {10.1109/iros.2016.7759644},
  abstract = {We present the localization approach for Astrobee, a new free-flying robot designed to navigate autonomously on the International Space Station (ISS). Astrobee will accommodate a variety of payloads and enable guest scientists to run experiments in zero-g, as well as assist astronauts and ground controllers. Astrobee will replace the SPHERES robots which currently operate on the ISS, whose use of fixed ultrasonic beacons for localization limits them to work in a 2 meter cube. Astrobee localizes with monocular vision and an IMU, without any environmental modifications. Visual features detected on a pre-built map, optical flow information, and IMU readings are all integrated into an extended Kalman filter (EKF) to estimate the robot pose. We introduce several modifications to the filter to make it more robust to noise, and extensively evaluate the localization algorithm.}
}

Soussan, R., Kumar, V., Coltin, B. and Smith, T. (2022). AstroLoc: An Efficient and Robust Localizer for a Free-Flying Robot . International Conference on Robotics and Automation. Source

BibTeX

@inproceedings{soussan2022astroloc,
  title = {AstroLoc: An Efficient and Robust Localizer for a Free-Flying Robot},
  author = {Soussan, Ryan and Kumar, Varsha and Coltin, Brian and Smith, Trey},
  booktitle = {International Conference on Robotics and Automation},
  pages = {4106-4112},
  year = {2022},
  doi = {10.1109/icra46639.2022.9811919},
  abstract = {We present AstroLoc, an efficient and robust monocular visual-inertial graph-based localization system used by the Astrobee free-flying robots onboard the International Space Station (ISS). We provide a novel localization system that limits the traditionally higher computation times for graph-based localization systems and enables the resource constrained Astrobee robots to benefit from their increased accuracy. We also introduce methods for handling cheirality issues for visual odometry and localization factors that further increase localization robustness. We evaluate the performance of AstroLoc on a dataset of ISS activities and show that it greatly improves pose, velocity, and IMU bias estimation accuracy while efficiently running in a limited computation environment. AstroLoc has improved the localization accuracy for the Astrobee robots on the ISS and has led to more successful and longer duration activities. While the AstroLoc system is tuned for the Astrobee robots, it can be configured for any resource constrained platform. The source code for AstroLoc is released to the public.}
}

Alexandrov, O., Barlow, J., Benavides, J., Bualat, M., Carlino, R., Coltin, B., Cortez, J., Daley, E., Feller, J., Flückiger, L., Fong, T., Fusco, J., Garcia Ruiz, R., Hamilton, K., Kanis, S., Katterhagen, A., Kim, Y., Love, J. F., McIntyre, M., McLachlan, B., Mora Vargas, A., Moratto, Z., Moreira, M., Morse, T., Orosco, H., Park, I.-W., Provencher, C., Sanchez, H., Sharif, K., Smith, E., Smith, T., Soussan, R., Symington, A., Talavera, R. O., To, V., Wheeler, D. and Yoo, J. (2026). Astrobee: Free-Flying Robots for the International Space Station . IEEE Transactions on Field Robotics. Source

BibTeX

@article{alexandrov2026astrobee,
  title = {Astrobee: Free-Flying Robots for the International Space Station},
  author = {Alexandrov, Oleg and Barlow, Jonathan and Benavides, Jose and Bualat, Maria and Carlino, Roberto and Coltin, Brian and Cortez, Jose and Daley, Earl and Feller, Jeffrey and Flückiger, Lorenzo and Fong, Terrence and Fusco, Jesse and Garcia Ruiz, Ruben and Hamilton, Kathryn and Kanis, Simeon and Katterhagen, Aric and Kim, Yunkyung and Love, John F. and McIntyre, Michael and McLachlan, Blair and Mora Vargas, Andres and Moratto, Zack and Moreira, Marina and Morse, Theodore and Orosco, Henry and Park, In-Won and Provencher, Christopher and Sanchez, Hugo and Sharif, Khaled and Smith, Ernest and Smith, Trey and Soussan, Ryan and Symington, Andrew and Talavera, Rafael Omar and To, Vinh and Wheeler, DW and Yoo, Jongwoon},
  journal = {IEEE Transactions on Field Robotics},
  year = {2026},
  url = {https://ntrs.nasa.gov/citations/20260001396},
  abstract = {The Astrobees are free-flying robots that operate inside the International Space Station (ISS) and were launched to the ISS
in 2019. Since then they have successfully performed hundreds of activities in space supporting almost two dozen separate
research projects. The robots were designed to overcome multiple challenges unique to the ISS environment, including safety,
upgradeability and maintainability, limited mass and computation, and unique localization challenges from lack of gravity and a
constantly changing environment. This article provides an overview of Astrobee, from hardware and software design to deployment
results and activities.}
}

Arora, A., El-Hariry, M. and Olivares-Mendez, M. (2025). Reinforcement Learning Based 6-DoF Maneuvers for Microgravity Intravehicular Docking: A Simulation Study with Int-Ball2 in ISS-JEM . arXiv. Source

BibTeX

@article{arora2025reinforcement,
  title = {Reinforcement Learning Based 6-DoF Maneuvers for Microgravity Intravehicular Docking: A Simulation Study with Int-Ball2 in ISS-JEM},
  author = {Arora, Aman and El-Hariry, Matteo and Olivares-Mendez, Miguel},
  journal = {arXiv},
  year = {2025},
  doi = {10.48550/arxiv.2512.13514},
  abstract = {Autonomous free-flyers play a critical role in intravehicular tasks aboard the International Space Station (ISS), where their precise docking under sensing noise, small actuation mismatches, and environmental variability remains a nontrivial challenge. This work presents a reinforcement learning (RL) framework for six-degree-of-freedom (6-DoF) docking of JAXA's Int-Ball2 robot inside a high-fidelity Isaac Sim model of the Japanese Experiment Module (JEM). Using Proximal Policy Optimization (PPO), we train and evaluate controllers under domain-randomized dynamics and bounded observation noise, while explicitly modeling propeller drag-torque effects and polarity structure. This enables a controlled study of how Int-Ball2's propulsion physics influence RL-based docking performance in constrained microgravity interiors. The learned policy achieves stable and reliable docking across varied conditions and lays the groundwork for future extensions pertaining to Int-Ball2 in collision-aware navigation, safe RL, propulsion-accurate sim-to-real transfer, and vision-based end-to-end docking.}
}

Bualat, M. G., Smith, T., Fong, T. W., Smith, E. E. and Wheeler, D. W. (2018). Astrobee: A New Tool for ISS Operations . SpaceOps Conference. Source

BibTeX

@inproceedings{bualat2018astrobee,
  title = {Astrobee: A New Tool for ISS Operations},
  author = {Bualat, Maria G. and Smith, Trey and Fong, Terrence W. and Smith, Ernest E. and Wheeler, D. W.},
  booktitle = {SpaceOps Conference},
  address = {Marseille, France},
  year = {2018},
  doi = {10.2514/6.2018-2517},
  abstract = {The Astrobee}
}

Bualat, M., Barlow, J., Fong, T., Provencher, C., Smith, T. and Zuniga, A. (2015). Astrobee: Developing a Free Flying Robot for the International Space Station . AIAA SPACE Conference, 20150018250. Source

BibTeX

@inproceedings{bualat2015astrobee,
  title = {Astrobee: Developing a Free Flying Robot for the International Space Station},
  author = {Bualat, Maria and Barlow, Jonathan and Fong, Terrence and Provencher, Christopher and Smith, Trey and Zuniga, Allison},
  booktitle = {AIAA SPACE Conference},
  number = {20150018250},
  institution = {NASA},
  year = {2015},
  doi = {10.2514/6.2015-4643},
  abstract = {Astronaut time will always be in short supply, consumables (e.g., oxygen) will always be limited, and some work will not be feasible, or productive, for astronauts to do manually. Free flyers offer significant potential to perform a great variety of tasks, include routine, repetitive or simple but long-duration work, such as conducting environment surveys, taking sensor readings or monitoring crew activities. The "Astrobee" project is developing a new free flying robot system suitable for performing Intravehicular Activity (IVA) work on the International Space Station (ISS). This paper will describe the Astrobee project objectives, initial design, concept of operations, and key challenges.\n\n\n}
}

Chapin, S., Stewart, K., Leontie, R. and Henshaw, C. G. (2025). Autonomous Planning In-space Assembly Reinforcement-learning free-flYer (APIARY) International Space Station Astrobee Testing . International Conference on Space Robotics. Source

BibTeX

@inproceedings{chapin2025bautonomous,
  title = {Autonomous Planning In-space Assembly Reinforcement-learning free-flYer (APIARY) International Space Station Astrobee Testing},
  author = {Chapin, Samantha and Stewart, Kenneth and Leontie, Roxana and Henshaw, Carl Glen},
  booktitle = {International Conference on Space Robotics},
  pages = {418-424},
  year = {2025},
  doi = {10.1109/isparo66239.2025.11436341},
  abstract = {The US Naval Research Laboratory’s (NRL’s) Autonomous Planning In-space Assembly Reinforcement-learning free-flYer (APIARY) experiment pioneers the use of reinforcement learning (RL) for control of free-flying robots in the zero-gravity (zero-G) environment of space. On Tuesday, May 27th 2025 the APIARY team conducted the first ever, to our knowledge, RL control of a free-flyer in space using the NASA Astrobee robot on-board the International Space Station (ISS). A robust 6-degrees of freedom (DOF) control policy was trained using an actor-critic Proximal Policy Optimization (PPO) network within the NVIDIA Isaac Lab simulation environment, randomizing over goal poses and mass distributions to enhance robustness. This paper details the simulation testing, ground testing, and flight validation of this experiment. This on-orbit demonstration validates the transformative potential of RL for improving robotic autonomy, enabling rapid development and deployment (in minutes to hours) of tailored behaviors for space exploration, logistics, and real-time mission needs.}
}

Saenz-Otero, A. (2000). The SPHERES Satellite Formation Flight Testbed: Design and Initial Control. Source

BibTeX

@mastersthesis{saenzotero2000spheres,
  title = {The SPHERES Satellite Formation Flight Testbed: Design and Initial Control},
  author = {Saenz-Otero, Alvar},
  school = {Massachusetts Institute of Technology},
  type = {M.Eng. thesis},
  year = {2000},
  url = {https://dspace.mit.edu/handle/1721.1/86730}
}

Fong, T., Micire, M., Morse, T., Park, E., Provencher, C., To, V., Wheeler, D. W., Mittman, D., Torres, R. J. and Smith, E. (2013). Smart SPHERES: a Telerobotic Free-Flyer for Intravehicular Activities in Space . AIAA SPACE Conference. Source

BibTeX

@inproceedings{fong2013smart,
  title = {Smart SPHERES: a Telerobotic Free-Flyer for Intravehicular Activities in Space},
  author = {Fong, Terrence and Micire, Mark and Morse, Ted and Park, Eric and Provencher, Chris and To, Vinh and Wheeler, D. W. and Mittman, David and Torres, R. Jay and Smith, Ernest},
  booktitle = {AIAA SPACE Conference},
  publisher = {American Institute of Aeronautics and Astronautics},
  address = {San Diego, California},
  year = {2013},
  doi = {10.2514/6.2013-5338},
  abstract = {Smart SPHERES is a prototype free-flying space robot based on the SPHERES platform. Smart SPHERES can be remotely operated by astronauts inside a spacecraft, or by mission controllers on the ground. We developed Smart SPHERES to perform a variety of intravehicular activities (IVA), such as operations inside the International Space Station (ISS). These IVA tasks include environmental monitoring surveys (radiation, sound levels, etc.), inventory, and mobile camera work. In this paper, we first discuss the motivation for free-flying space robots. We then describe the development of the Smart SPHERES prototype, including avionics, software, and data communications. Finally, we present results of initial flight tests on-board the ISS.}
}

Tweddle, B. E. (2013). Computer Vision-Based Localization and Mapping of an Unknown, Uncooperative and Spinning Target for Spacecraft Proximity Operations. Source

BibTeX

@phdthesis{tweddle2013computer,
  title = {Computer Vision-Based Localization and Mapping of an Unknown, Uncooperative and Spinning Target for Spacecraft Proximity Operations},
  author = {Tweddle, Brent E.},
  school = {Massachusetts Institute of Technology},
  type = {Ph.D. thesis},
  year = {2013},
  url = {https://dspace.mit.edu/handle/1721.1/85693}
}

Hilstad, M. O. (2002). A Multi-Vehicle Testbed and Interface Framework for the Development and Verification of Separated Spacecraft Control Algorithms. Source

BibTeX

@mastersthesis{hilstad2002multi,
  title = {A Multi-Vehicle Testbed and Interface Framework for the Development and Verification of Separated Spacecraft Control Algorithms},
  author = {Hilstad, Mark O.},
  school = {Massachusetts Institute of Technology},
  type = {S.M. thesis},
  year = {2002},
  url = {https://dspace.mit.edu/handle/1721.1/16738}
}

Fredrickson, S., Duran, S., Braun, A., Straube, T. and Mitchell, J. (2006). AERCam Autonomy: Intelligent Software Architecture for Robotic Free Flying Nanosatellite Inspection Vehicles . Space. Source

BibTeX

@inproceedings{fredrickson2006aercam,
  title = {AERCam Autonomy: Intelligent Software Architecture for Robotic Free Flying Nanosatellite Inspection Vehicles},
  author = {Fredrickson, Steven and Duran, Steve and Braun, Angela and Straube, Timothy and Mitchell, Jennifer},
  booktitle = {Space},
  year = {2006},
  doi = {10.2514/6.2006-7392},
  abstract = {The NASA Johnson Space Center has developed a nanosatellite-class Free Flyer intended for future external inspection and remote viewing of human spacecraft.The Miniature Autonomous Extravehicular Robotic Camera (Mini AERCam) technology demonstration unit has been integrated into the approximate form and function of a flight system.The spherical Mini AERCam Free Flyer is 7.5 inches in diameter and weighs approximately 10 pounds, yet it incorporates significant additional capabilities compared to the 35-pound, 14inch diameter AERCam Sprint that flew as a Shuttle flight experiment in 1997.Mini AERCam hosts a full suite of miniaturized avionics, instrumentation, communications, navigation, power, propulsion, and imaging subsystems, including digital video cameras and a high resolution still image camera.The vehicle is designed for either remotely piloted operations or supervised autonomous operations, including automatic stationkeeping, pointto-point maneuvering, and waypoint tracking.The Mini AERCam Free Flyer is accompanied by a sophisticated control station for command and control, as well as a docking system for automated deployment, docking, and recharge at a parent spacecraft.Free Flyer functional testing has been conducted successfully on both an airbearing table and in a six-degree-of-freedom closed-loop orbital simulation with avionics hardware in the loop.Mini AERCam aims to provide beneficial on-orbit views that cannot be obtained from fixed cameras, cameras on robotic manipulators, or cameras carried by crewmembers during extravehicular activities (EVA's).On Shuttle or International Space Station (ISS), for example, Mini AERCam could support external robotic operations by supplying orthogonal views to the intravehicular activity (IVA) robotic operator, supply views of EVA operations to IVA and/or ground crews monitoring the EVA, and carry out independent visual inspections of areas of interest around the spacecraft.To enable these future benefits with minimal impact on IVA operators and ground controllers, the Mini AERCam system architecture incorporates intelligent systems attributes that support various autonomous capabilities.1) A robust command sequencer enables task-level command scripting.Command scripting is employed for operations such as automatic inspection scans over a region of interest, and operator-hands-off automated docking.2) A system manager built on the same expert-system software as the command sequencer provides detection and smartresponse capability for potential system-level anomalies, like loss of communications between the Free Flyer and control station.3) An AERCam dynamics manager provides nominal and off-nominal management of guidance, navigation, and control (GN&C) functions.It is employed for safe trajectory monitoring, contingency maneuvering, and related roles.This paper will describe these architectural components of Mini AERCam autonomy, as well as the interaction of these elements with a human operator during supervised autonomous control.}
}

Fredrickson, S. E., Abbott, L. W., Duran, S., Jochim, J. D., Studak, J. W., Wagenknecht, J. D. and Williams, N. M. (2003). Mini AERCam: Development of a Free-Flying Nanosatellite Inspection Robot . Space. Source

BibTeX

@inproceedings{fredrickson2003mini,
  title = {Mini {AERCam}: Development of a Free-Flying Nanosatellite Inspection Robot},
  author = {Fredrickson, Steven E. and Abbott, Larry W. and Duran, Steve and Jochim, J. David and Studak, J. William and Wagenknecht, Jennifer D. and Williams, Nichole M.},
  booktitle = {Space},
  volume = {5088},
  pages = {97-111},
  publisher = {SPIE},
  year = {2003},
  doi = {10.1117/12.498108},
  abstract = {The Engineering Directorate of NASA Johnson Space Center has developed a nanosatellite-class free-flyer intended for future external inspection and remote viewing of human spaceflight activities, including International Space Station (ISS) operations. The Miniature Autonomous Extravehicular Robotic Camera (Mini AERCam) technology demonstration unit has been integrated into the approximate form and function of a flight system. The spherical Mini AERCam free flyer is 7.5 inches in diameter and weighs approximately 10 pounds, yet it incorporates significant additional capabilities compared to the 35 pound, 14 inch AERCam Sprint that flew as a Shuttle flight experiment in 1997. Mini AERCam hosts a full suite of miniaturized avionics, instrumentation, communications, navigation, imaging, power, and propulsion subsystems, including two digital video cameras and a high resolution still image camera. The vehicle is designed for either remotely piloted operations or supervised autonomous operations including automatic stationkeeping and point-to-point maneuvering. Free-flyer testing has been conducted on an air-bearing table and in a six degree-of-freedom closed-loop orbital simulation. The orbital simulation models the three-dimensional dynamics of the free-flyer in proximity to the ISS, and produces corresponding God's eye views and simulated free-flyer camera views. A high-fidelity simulation is achieved by directly interfacing to free-flyer thruster driver signals, emulating the MEMS gyro responses in hardware, and using the "truth" state to drive a GPS signal generator connected to the free-flyer GPS receiver.}
}

Turchetti, F., Ekal, M., Lii, N. Y. and Roa, M. A. (2024). Analysis of Intra-Vehicular Robotic Free-Flyers and Their Manipulation Capabilities . International Astronautical Congress, IAC-24-A2.5.10. Source

BibTeX

@inproceedings{turchetti2024analysis,
  title = {Analysis of Intra-Vehicular Robotic Free-Flyers and Their Manipulation Capabilities},
  author = {Turchetti, Federico and Ekal, Monica and Lii, Neal Y. and Roa, Maximo A.},
  booktitle = {International Astronautical Congress},
  number = {IAC-24-A2.5.10},
  year = {2024},
  url = {https://elib.dlr.de/208576/}
}

Hirano, D., Mitani, S., Watanabe, K., Nishishita, T., Yamamoto, T. and Yamaguchi, S. P. (2025). Int-Ball2: On-Orbit Demonstration of Autonomous Intravehicular Flight and Docking for Image Capturing and Recharging . IEEE Robotics & Automation Magazine, 3. Source

BibTeX

@article{hirano2025intball2,
  title = {{Int-Ball2}: On-Orbit Demonstration of Autonomous Intravehicular Flight and Docking for Image Capturing and Recharging},
  author = {Hirano, Daichi and Mitani, Shinji and Watanabe, Keisuke and Nishishita, Taisei and Yamamoto, Tatsuya and Yamaguchi, Seiko P.},
  journal = {IEEE Robotics & Automation Magazine},
  volume = {32},
  number = {3},
  pages = {76--87},
  year = {2025},
  doi = {10.1109/mra.2024.3505776},
  abstract = {This article presents the system architecture and the orbital demonstration results of the Int-Ball2, a free-flying camera robot developed by the Japan Aerospace Exploration Agency (JAXA). The purpose of the Int-Ball2 project is to assist astronauts and reduce their workload in the International Space Station (ISS). This robot is an upgrade from the first Int-Ball, enhancing the propulsion subsystem for greater maneuverability and adding a new docking station (DS) for autonomous battery recharging. This study performed comprehensive ground tests for autonomous maneuvering and docking, employing a combination of a fully software-based simulator, a hardware-in-the-loop (HIL) simulator, and a planar air-bearing facility. After a successful launch to the ISS, the Int-Ball2 demonstrated its ability to work in microgravity without relying on astronaut support. The results obtained from ground and orbital tests underscored the effectiveness of our system design and ground verification approach. Further, we present key technologies essential for the Int-Ball2’s successful implementation on board the ISS. We expect the insights from this project to be invaluable to future missions involving free-flying robots in microgravity.}
}

Dinkel, H., Di, J., Santos, J., Albee, K., Borges, P., Moreira, M., Alexandrov, O., Coltin, B. and Smith, T. (2023). Multi-Agent 3D Map Reconstruction and Change Detection in Microgravity with Free-Flying Robots . International Astronautical Congress, IAC-23,D1,6,1,x78669. Source

BibTeX

@inproceedings{dinkel2023multi,
  title = {Multi-Agent 3D Map Reconstruction and Change Detection in Microgravity with Free-Flying Robots},
  author = {Dinkel, Holly and Di, Julia and Santos, Jamie and Albee, Keenan and Borges, Paulo and Moreira, Marina and Alexandrov, Oleg and Coltin, Brian and Smith, Trey},
  booktitle = {International Astronautical Congress},
  number = {IAC-23,D1,6,1,x78669},
  publisher = {JPL Open Repository},
  year = {2023},
  doi = {10.48577/jpl.lbsjso},
  abstract = {Assistive free-flyer robots autonomously caring for future crewed outposts---such as NASA’s Astrobee robots on the International Space Station (ISS)---must be able to detect day-to-day interior changes to track inventory, detect and diagnose faults, and monitor the outpost status. This work presents a framework for multi-agent cooperative mapping and change detection to enable robotic maintenance of space outposts. One agent is used to reconstruct a 3D model of the environment from sequences of images and corresponding depth information. Another agent is used to scan the environment for inconsistencies against the 3D model. Change detection is validated using real image and pose data collected by Astrobee robots on Earth in a test environment and in space on the ISS. This work outlines the objectives, requirements, and algorithmic modules for the multi-agent reconstruction system, including recommendations for its use aboard future microgravity outposts.}
}

Yang, E., Hwu, S. U., Lansdowne, C., Boster, J. P. and deSilva, K. (2024). Wi-Fi Signal Survey of the International Space Station by Autonomous Free-Flying Robot . IEEE International Conference on Wireless for Space and Extreme Environments, 20240014199. Source

BibTeX

@inproceedings{yang2024wi,
  title = {Wi-Fi Signal Survey of the International Space Station by Autonomous Free-Flying Robot},
  author = {Yang, Everest and Hwu, Shian U. and Lansdowne, Chatwin and Boster, John P. and deSilva, Kanishka},
  booktitle = {IEEE International Conference on Wireless for Space and Extreme Environments},
  number = {20240014199},
  pages = {238-243},
  institution = {NASA},
  year = {2024},
  doi = {10.1109/wisee61249.2024.10850420},
  abstract = {This paper analyzes Wi-Fi signal propagation inside pressurized modules of the International Space Station (ISS). Various flight datasets were collected by the Astrobee fleet of autonomous free-flying robots developed by the NASA Ames Research Center (ARC). The enclosed module spaces of the ISS pose unique challenges for Wi-Fi signal propagation, and understanding these challenges is crucial for optimizing connectivity in such environments. The survey data presented are valuable for optimizing wireless access point (WAP) locations and conditions. A computational method is developed to analyze the wireless communication system, RF coverage, and module compositions. The method is both rigorous and practical for assessing the ISS wireless system performance. The computational tools in this paper serve as a complementary approach to direct, on-orbit measurements, which is a complicated and expensive task. In some operational scenarios, ground measurements may be intractable due to the size of the spacecraft. Results from this research will aid future space missions and commercial spacecraft, providing insights into optimal WAP placements and robust wireless communication systems.}
}

Miller, D. L. (2015). Development of Resource-Constrained Sensors and Actuators for In-Space Satellite Docking and Servicing. Source

BibTeX

@mastersthesis{miller2015development,
  title = {Development of Resource-Constrained Sensors and Actuators for In-Space Satellite Docking and Servicing},
  author = {Miller, Duncan L.},
  school = {Massachusetts Institute of Technology},
  type = {S.M. thesis},
  year = {2015},
  url = {https://dspace.mit.edu/handle/1721.1/98805}
}

Mohan, S., Saenz-Otero, A., Nolet, S., Miller, D. W. and Sell, S. (2009). SPHERES Reconfigurable Framework and Control System Design for Autonomous Assembly . AIAA Guidance, Navigation, and Control Conference. Source

BibTeX

@inproceedings{mohan2009spheres,
  title = {SPHERES Reconfigurable Framework and Control System Design for Autonomous Assembly},
  author = {Mohan, Swati and Saenz-Otero, Alvar and Nolet, Simon and Miller, David W. and Sell, Steven},
  booktitle = {AIAA Guidance, Navigation, and Control Conference},
  year = {2009},
  doi = {10.2514/6.2009-5978},
  abstract = {Reconfigurable control system design is a key component for enabling autonomous on-orbit assembly. Current research on reconfigurable control systems focuses on adapting to failures. However, for assembly scenarios, the reconfiguration is necessitated by changing mass and stiffness properties. This paper provides a brief description of existing reconfigurable control system technology and develops a framework to incorporate reconfiguration into an existing baseline system to account for mass property variations. The reconfigurable control system framework has been developed and implemented using the SPHERES (Synchronized Position Hold Engage Reorient Experimental Satellites) testbed as the baseline system. The framework highlights the elements that need to be updated, introduces a variable p that captures the configuration, and details the updates necessary in the key algorithms to calculate the model on-line using p. Results are presented from the implementation on the SPHERES, focusing on the reconfigurable estimator. Plans are presented for an integrated assembly test that demonstrates the maintenance of stability, fuel efficiency, and accuracy throughout configuration changes that occur during assembly.}
}

Hilton, A. R. (2015). A Performance-Driven Experiment Framework for Space Technology Development Using the International Space Station. Source

BibTeX

@mastersthesis{hilton2015performance,
  title = {A Performance-Driven Experiment Framework for Space Technology Development Using the International Space Station},
  author = {Hilton, Andrew R.},
  school = {Massachusetts Institute of Technology},
  type = {S.M. thesis},
  year = {2015},
  url = {https://dspace.mit.edu/handle/1721.1/98559}
}

Chen, A. (2002). Propulsion System Characterization for the SPHERES Formation Flight and Docking Testbed. Source

BibTeX

@mastersthesis{chen2002propulsion,
  title = {Propulsion System Characterization for the SPHERES Formation Flight and Docking Testbed},
  author = {Chen, Allen},
  school = {Massachusetts Institute of Technology},
  type = {S.M. thesis},
  year = {2002},
  url = {https://dspace.mit.edu/handle/1721.1/16736}
}

(2021). Airbus: CIMON-2 makes its successful debut on the ISS. airbus.com/en/newsroom/press-releases/2020-04-cimon-2-makes-its-succe...

BibTeX

@misc{airbuscimon,
  title = {Airbus: CIMON-2 makes its successful debut on the ISS},
  organization = {airbus.com},
  year = {2021},
  url = {https://www.airbus.com/en/newsroom/press-releases/2020-04-cimon-2-makes-its-successful-debut-on-the-iss}
}

Soussan, R., Moreira, M., Coltin, B. and Smith, T. (2025). AstroLoc2: Fast Sequential Depth-Enhanced Localization for Free-flying Robots . IEEE International Conference on Robotics and Automation. Source

BibTeX

@inproceedings{soussan2025astroloc2,
  title = {AstroLoc2: Fast Sequential Depth-Enhanced Localization for Free-flying Robots},
  author = {Soussan, Ryan and Moreira, Marina and Coltin, Brian and Smith, Trey},
  booktitle = {IEEE International Conference on Robotics and Automation},
  pages = {14595-14602},
  address = {Atlanta, Georgia},
  year = {2025},
  doi = {10.1109/icra55743.2025.11128673},
  abstract = {We present AstroLoc2, a monocular and time-offlight (ToF) visual-inertial graph-based localizer used by the Astrobee free-flying robots on the International Space Station (ISS). AstroLoc2 sequentially performs odometry and absolute localization in a single process to decouple map noise from velocity and IMU bias estimation and run efficiently on resource constrained platforms. It improves monocular visual-inertial odometry robustness by adding ToF correspondence factors and uses adaptive map-matching to increase image registration reliability in dynamic environments while preserving fast matching in static ones. We evaluate the performance of AstroLoc2 on a public dataset of 10 ISS activities and show that it improves localization accuracy by 16 % and success rates by 5.5 % while maintaining a faster runtime than leading methods. AstroLoc2 has enabled the Astrobee robots to perform higher precision maneuvers in changing environments on the ISS. It can be configured for other limited computation platforms and we release the source code to the public.}
}

Miller, I. D., Soussan, R., Coltin, B., Smith, T. and Kumar, V. (2022). Robust Semantic Mapping and Localization on a Free-Flying Robot in Microgravity . International Conference on Robotics and Automation. Source

BibTeX

@inproceedings{miller2022robust,
  title = {Robust Semantic Mapping and Localization on a Free-Flying Robot in Microgravity},
  author = {Miller, Ian D. and Soussan, Ryan and Coltin, Brian and Smith, Trey and Kumar, Vijay},
  booktitle = {International Conference on Robotics and Automation},
  pages = {4121-4127},
  year = {2022},
  doi = {10.1109/icra46639.2022.9811862},
  abstract = {We propose a system that uses semantic object detections to localize a microgravity free-flyer. Many applications require absolute localization in a known reference frame, such as the execution of waypoint trajectories defined by human operators. Classical geometric methods build a map of point features, which may not be able to be associated after lighting or environmental changes. By contrast, semantics remain invariant to changes up to the robustness of the detection algorithm and motion of the semantic objects. In this work, we describe our approaches for both offline semantic map generation as well as online localization against a semantic map, intended to run in real-time on the robot. We additionally demonstrate how our semantic localizer outperforms image-feature matching in some cases, and show the robustness of the algorithm to environmental changes. Crucially, we show in our experiments that when semantics are used to supplement point features, localization is always improved. To our knowledge, these experiments demonstrate the first use of learned semantics for localization on a free-flying robot in microgravity.}
}

Welde, J., Rao, N., Kunapuli, P., Jayaraman, D. and Kumar, V. (2024). Leveraging Symmetry to Accelerate Learning of Trajectory Tracking Controllers for Free-Flying Robotic Systems . IEEE International Conference on Robotics and Automation. Source

BibTeX

@inproceedings{welde2024leveraging,
  title = {Leveraging Symmetry to Accelerate Learning of Trajectory Tracking Controllers for Free-Flying Robotic Systems},
  author = {Welde, Jake and Rao, Nishanth and Kunapuli, Pratik and Jayaraman, Dinesh and Kumar, Vijay},
  booktitle = {IEEE International Conference on Robotics and Automation},
  pages = {15121-15127},
  year = {2024},
  doi = {10.1109/icra55743.2025.11127455},
  abstract = {Tracking controllers enable robotic systems to accurately follow planned reference trajectories. In particular, reinforcement learning (RL) has shown promise in the synthesis of controllers for systems with complex dynamics and modest online compute budgets. However, the poor sample efficiency of RL and the challenges of reward design make training slow and sometimes unstable, especially for high-dimensional systems. In this work, we leverage the inherent Lie group symmetries of robotic systems with a floating base to mitigate these challenges when learning tracking controllers. We model a general tracking problem as a Markov decision process (MDP) that captures the evolution of both the physical and reference states. Next, we prove that symmetry in the underlying dynamics and running costs leads to an MDP homomorphism, a mapping that allows a policy trained on a lower-dimensional “quotient” MDP to be lifted to an optimal tracking controller for the original system. We compare this symmetry-informed approach to an unstructured baseline, using Proximal Policy Optimization (PPO) to learn tracking controllers for three systems: the Particle (a forced point mass), the Astrobee (a fully-actuated space robot), and the Quadrotor (an underactuated system). Results show that a symmetry-aware approach both accelerates training and reduces tracking error at convergence.}
}

Albee, K., Sternberg, D. C., Hansson, A., Schwartz, D., Majumdar, R. and Jia-Richards, O. (2025). Architecting Autonomy for Safe Microgravity Free-Flyer Inspection . IEEE Aerospace Conference. Source

BibTeX

@inproceedings{albee2025architecting,
  title = {Architecting Autonomy for Safe Microgravity Free-Flyer Inspection},
  author = {Albee, Keenan and Sternberg, David C. and Hansson, Alexander and Schwartz, David and Majumdar, Ritwik and Jia-Richards, Oliver},
  booktitle = {IEEE Aerospace Conference},
  publisher = {JPL Open Repository},
  year = {2025},
  doi = {10.48577/jpl.esh2bl}
}

Stewart, K. H., Chapin, S., Leontie, R. and Henshaw, C. G. (2025). Crossing the Sim2Real Gap Between Simulation and Ground Testing to Space Deployment of Autonomous Free-flyer Control . arXiv. Source

BibTeX

@article{stewart2025crossing,
  title = {Crossing the Sim2Real Gap Between Simulation and Ground Testing to Space Deployment of Autonomous Free-flyer Control},
  author = {Stewart, K. H. and Chapin, Samantha and Leontie, Roxana and Henshaw, Carl Glen},
  journal = {arXiv},
  year = {2025},
  doi = {10.1109/isparo66239.2025.11436888},
  abstract = {Reinforcement learning (RL) offers transformative potential for robotic control in space. We present the first on-orbit demonstration of RL-based autonomous control of a free-flying robot, the NASA Astrobee, aboard the International Space Station (ISS). Using NVIDIA's Omniverse physics simulator and curriculum learning, we trained a deep neural network to replace Astrobee's standard attitude and translation control, enabling it to navigate in microgravity. Our results validate a novel training pipeline that bridges the simulation-to-reality (Sim2Real) gap, utilizing a GPU-accelerated, scientific-grade simulation environment for efficient Monte Carlo RL training. This successful deployment demonstrates the feasibility of training RL policies terrestrially and transferring them to space-based applications. This paves the way for future work in In-Space Servicing, Assembly, and Manufacturing (ISAM), enabling rapid on-orbit adaptation to dynamic mission requirements.}
}