Small Space Robots for In-orbit Operations

Lead Research Organisation: University of Surrey
Department Name: Surrey Space Centre Academic

Abstract

There is an increasing interest to deploy robotic spacecraft for a range of in-orbit operations; including assembly and servicing missions. These involve building large structures, the in-situ repair of functional spacecraft approaching the end of their mission, and the cleaning-up of space debris. Previous technology demonstrations, such as Orbital Express and the Experimental Test Satellite use medium to large robotic spacecraft, and the current trend in the industry is towards building larger robotic spacecraft to handle the ever-growing size of target systems. However, few studies have looked into the feasibility of a low cost, downsized alternative. This project is tailored to look at the utility of a small spacecraft and how two identical systems can dock and reconfigure into a conjoined twin robotic system. The first generation of this small robotic spacecraft will be hereafter referred to as "Twin-Sat". One of the advantages of Twin-Sat is that it will result in creating a more stable and bigger base spacecraft with dual arms for executing heavy duty tasks, therefore increasing its capability. The other advantage is that the conjoined twins would be capable of decoupling themselves into individual spacecraft, allowing them to handle small-scale tasks that require multiple agents.
The key objectives of this research are to conduct a system level electro-mechanical design iteration to determine the optimum sizing of the single system as well as to model the dynamic characteristics of two identical, docked space robots; "Twin-Sat". Further to this the dexterity and payload capacity of the robotic arm, with and without in-built compliance will be assessed. This research will also deliver a detailed engineering model of the space robot, control algorithms for Twin-Sat under different operating conditions, and size recommendations based on extensive simulation and experimental validation. This project is partly funded by Surrey Satellite Technology Ltd. (SSTL) and began in October 2018.

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/N509772/1 30/09/2016 29/09/2021
2116510 Studentship EP/N509772/1 30/09/2018 30/03/2022 LUCY JACKSON
EP/R513350/1 30/09/2018 29/09/2023
2116510 Studentship EP/R513350/1 30/09/2018 30/03/2022 LUCY JACKSON
 
Description While robotic technologies have been long-serving in space, and are still an active and ever-growing field of research, very few autonomous robotics have been used beyond the ISS and surrounding area. Satellite mounted manipulators would allow more ambitious tasks to be carried out in a safer and more timely manner, by limiting the need for astronaut intervention during task execution. This would facilitate missions such as active debris removal and on-orbit assembly - two areas that are attracting a lot of government and commercial interest.
Limited research has been done in assessing the practical challenges involved in launching and operating a small space robot. We applied classical and modern deep learning optimisation techniques to the task of designing and controlling a small space robot suitable for on-orbit operations. Our research not only provides a method suitable for re-implementation to other tasks in which the simultaneous design of both hardware and software is necessary but also knowledge as to the appropriate sizing of a satellite mounted manipulator. The design of such a system is a challenging one due to the phenomenon of dynamic coupling which is when the base of the satellite moves in reaction to every motion of the arm - this does not occur when operating on Earth under the effects of gravity. No prior research has quantified the necessary mass to length ratio of such systems. In addition we show that deep learning (in particular reinforcement learning) is a feasible and suitable option for use in the space robotics industry where at present it is rarely used. In particular we make it applicable through the development of a novel algorithm that allows a single controller to operate on a range of semi-identical agents. This is particularly suited to space robotics since development and integration of space qualified systems is particularly arduous, and qualifying a single system is much more desirable than many.
Exploitation Route The area of space robotics is fast maturing with the recent interest in active debris removal. The outcome of our design exercise maybe used directly as a starting point for future groups to develop and design a satellite mounted robotic arm. If not used directly to design such a system, our methodology could be applied to any software/hardware design process in which the physical system and control are heavily interlinked. This is because our algorithm showed to enhance the final performance of systems design in this way over conventional methods. In addition to this, our novel control algorithm could be used in any setting in which it is desirable to control a range of semi-identical agents with a single network. This might include assembly agents with varying link lengths or cars with different sized wheels or gearing systems.
Sectors Other
 
Description Our findings stemmed from a collaboration between SSTL and CVSSP, this has given both partners a chance to think about how machine learning could be used in space. In addition to this it has provided SSTL with a top level knowledge on robotic satellites that can be used in trade-offs for future projects. In addition to this, there is potential for the output design to directly inform the hardware and software design of future robotic satellite missions.
First Year Of Impact 2022
Sector Environment,Other
Impact Types Policy & public services