Ottawa’s Obruta Space Solutions is close to successfully concluding a year-long series of tests of their autonomous orbital docking technology. The tests have been conducted aboard the International Space Station over the past year, using the station’s “Astrobee” drones, and the company is anticipating that the final test will be happening sometime before the end of the year.
Obruta Co-founder and CEO, Kevin Stadnyk, discussed the tests and Obruta’s plans going forward in an interview with SpaceQ.
Obruta and the RPOD
Obruta, as discussed in previous SpaceQ coverage, was originally focused on debris removal, but pivoted to developing RPOD (Rendezvous, Proximity Operations, and Docking) technology instead. Their RPOD kits, which Obruta describes as a “turnkey solution to equip spacecraft for safe, reliable, and autonomous orbital docking,” are intended to give spacecraft the ability to maneuver and dock with each other, even when they wouldn’t have had that capability before. This opens the door for in-orbit refueling, servicing, orbit correction, and other operations that would have been difficult-to-impossible previously.
Stadnyk said that the company is aiming to “enable the broader in-space economy by supplying this capability, allowing teams to get to market much quicker than they would if they developed it in house.”
The Canadian Space Agency (CSA) provided two STDP awards to Obruta, stating they were intended to help with Obruta’s development of “Computer Vision for Autonomous Spacecraft Docking” and with “Guidance, Navigation, and Control for Autonomous Spacecraft Docking.” This round of tests, as mentioned in that earlier coverage, was also thanks to an 2022 National Laboratory Research Announcement (NLRA) award, in partnership with the American company GeoJump.
Testing on the ISS
After the first set of tests in February, Obruta announced the test on LinkedIn, noting that it was “a major milestone in the development of Obruta’s RPOD kit.” Two other testing events have happened since then, according to Stadnyk, with one more test coming sometime after Christmas but (hopefully) before the end of the year. This fourth set of tests will conclude the testing series aboard the ISS.
All of the tests were conducted using the Astrobees, which are autonomous drones owned by NASA’s Ames Research Center (or “NASA Ames”) used for testing aboard the ISS. Stadnyk said that they are “floating robotics platforms inside the GEM module of the ISS” that have six degrees of freedom, much like other spacecraft, as well as optics and “a small compute platform.” Obruta used two of the Astrobees to “fly a myriad of guidance paths, testing out different algorithms and testing out different control controllers for the overall software system.”
The first tests in February “looked at flying some specific guidance paths,” said Stadnyk, “and seeing the performance on orbit relative to our ground-based modelling.” In late February they had a second opportunity to use the Astrobees, and this time they tested opportunities for “harnessing the orbital effects and gravitational effects on orbit” for maneuvering the drones. Stadnyk said that “you can take advantage of those inside the space station just as you can outside of them, and we were looking at harnessing those effects for testing more fuel efficient and optimized trajectories.”
While it would seem strange to test orbital dynamics aboard the ISS, it makes sense; disconnected objects aboard the ISS are all in orbit around the Earth, just at the same speed and in the same location. Two Astrobees maneuvering around inside the ISS would behave much like two spacecraft maneuvering around outside of it, if they were in proximity and in the same orbit, which is similar to the spacecraft docking maneuvers that Obruta is focused on. All the slow-down-to-speed-up weirdness of orbital mechanics is in play.
Stadnyk described it as “sort of like a boat on the ocean taking advantage of the currents to help move you, even when you might not have all of the wind in your sails.”
In the third session in June, they moved beyond maneuvering to actual docking. Stadnyk noted that the Astrobees can’t actually dock with each other, so they worked with the NASA Ames team on the problem. They were eventually approved to, he said, “do a sort of bump-together of the Astrobees,” which lets them simulate the docking, “coming in at the appropriate speeds and angles and attitude rates.” They were able to show, he said, “that our trajectories could both bring the Astrobees [together] from a faraway point up to very close, safely perform proximity operations,” and then approach and bump the two Astrobees together.
Stadnyk said that, having completed all three sets of tests, “one of the most surprising things was that things were performing as expected [and] as designed.” While they did have some debugging to do for the software, and there were “initial difficulties integrating everything together with the Astrobee platforms,” their on-orbit tests matched what they’d anticipated based on terrestrial testing. He said that the tests “were able to successfully validate that our work on the ground was in fact viable in the microgravity and orbital environment.”
The final autonomous test
The upcoming final test may be somewhat more difficult. These tests will focus on bringing everything together and testing fully-autonomous Astrobee docking.
In the earlier tests, they weren’t able to test or train their AI-based computer vision — the hardware on the Astrobees wasn’t powerful enough to run the ML algorithms necessary to do the job. After the early tests, Stadnyk said that they “were able to work with the NASA Ames team to set up a new system on the Astrobee [to collect data] for training our vision based models.” They succeeded, and he said “we were actually the first team to run AI onboard the Astrobee in this way.” They’ve shared that information with other Astrobee users, and now those other users will be able to implement these systems in their own testing.
Thanks to the Astrobees’ collected training data, Stadnyk said, they’ve “validated the control and guidance systems on the ground,” and “were able to train vision-based AI models on the ground and test [them] in isolation.” So, in this final test, they will be “putting it all together, such that the Astrobee will be fully autonomous,” using “real time, computer vision-based relative navigation.”
The Astrobee, Stadnyk said, will “navigate visually within its environment,” charting an optimal path to the other Astrobee.” Then it will “control along that guidance path to achieve docking at its final destination.”
Stadnyk called it a “full end to end mission, putting everything we’ve been testing and developing together.” It will require some additional hardware on the Astrobee, and part of the reason this has taken a while is that they’ll need to add a GPU-like parallel processor to the Astrobee in order to give it the additional compute power needed for the AI. But once it’s done, if the test is successful, Obruta will have validated their toolkit in orbit.
Fundraising and further testing
When asked about what was next, Stadnyk said that Obruta is planning on fundraising for further development. He said that they are “actively fundraising to support the next tranche of our developments,” and are “looking forward to a new project under the CSA’s STDP program, where we will be continuing the development of our software.” In the new year, he said, “we are looking towards getting hardware engineering model units developed and readily available for customer initial testing and integration with their spacecraft.”
Obruta sees a lot of opportunity in the current market; investors want to see results and a quicker move to market, and a turnkey solution for docking can help space companies get there faster.
And, yes, they’re looking to do more orbital testing as well. Stadnyk said that “we are looking towards a full-scale flight mission to build upon the heritage of this Astrobee testing,” so that they can “fully validate the RPOD system in a real, full-size operational space environment.” At the moment the plan is to integrate with a larger spacecraft, then perform the RPO maneuvers with a small CubeSat deployed from the host spacecraft. They said that that would allow them to “run this mission in a similar manner to how we would be interfacing with customers in the long run.”