IEEE - Aerospace and Electronic Systems - June 2022 - 38

Feature Article:
DOI. No. 10.1109/MAES.2021.3089252
Reinforcement-Learning-Based Task Planning for
Self-Reconfiguration of Cellular Satellites
Yizhai Zhang , Northwestern Polytechnical University, Xi'an 710072, China,
and National Key Laboratory of Aerospace Flight Dynamics, Xi'an 710072,
China
Wenhui Wang, Shanghai Institute of Space Propulsion, Shanghai 201112,
China
Piaoqi Zhang and Panfeng Huang , Northwestern Polytechnical University,
Xi'an 710072, China, and National Key Laboratory of Aerospace Flight
Dynamics, Xi'an 710072, China
INTRODUCTION
With an increase in large-scale space explorations, more
spacecraft and satellites will be launched. However, traditional
design and operating modes of space systems do not
satisfy all future requirements. For example, future space
systems are expected to have low cost, rapid response,
high serviceability, multiple purposes, and even change
functions on-orbit for unplanned missions [1]-[3].
Cellular satellites represent a class of novel and promising
satellites and have the potential to fundamentally change
the design and operatingmodes ofspace systems [4], [5]. Cellular
satellites usually consist of many small cubic-like unit
cells. All unit cells are standardized, physically separated,
functionally independent, and interconnected through standard
interfaces. The standardized multifunctional interface
can not only enable mechanical connection, but also provide
high-speed optical data communication, electric transmission,
and thermal exchange. Hideyuki Tanaka first proposed
the concept of cellular satellites in 2006 [6]. Recently, much
progress has been made in the studies of cellular satellites.
Authors' current addresses: Y. Zhang, P. Zhang, P.
Huang, Northwestern Polytechnical University, Xi'an
710072, China; National Key Laboratory of Aerospace
Flight Dynamics, Xi'an 710072, China (e-mail: zhangyizhai@nwpu.edu.cn,
pqzhang@mail.nwpu.edu.cn,
pfhuang@nwpu.edu.cn). W. Wang, Shanghai Institute
of Space Propulsion, Shanghai 201112, China (e-mail:
359485967@qq.com).
Manuscript received November 8, 2020, revised May
10, 2021; accepted May 31, 2021, and ready for
publication June 11, 2021.
Review handled by Tai-Hoon Kim.
0885-8985/21/$26.00 ß 2021 IEEE
38
Deutsches Zentrum fur L€uft-und Raumfahrt, GER (DLR)
developed the Intelligent Building Blocks for On-Orbit-Satellite
Servicing (iBOSS) plan, demonstrating many cell prototypes
and conducting ground experiments [7]-[9] (see
Figure 1). In the iBOSS plan, the satellite is divided into standardized
building blocks for on-orbit servicing. The relevant
subsystem components are installed inside each building
block. The Defense Advanced Research Projects Agency,
USA (DARPA) released the " satlet " concept in the Phoenix
Program [10]. Each satlet performs a part ofthe overall functions,
and the satlets are interconnected via hardware and
software. We also developed two generations of cell prototypes
for cellular satellites and began to investigate the selfreconfiguration
problems at Northwestern Polytechnical University,
China (NWPU) [11]-[13] (see Figure 2). For the cellular
satellites at NWPU, one class of unit cells are strict
cubes with six identical surfaces, such as controller cells and
computation cells. The other class represents the cells with
one or two specialized sides, such as thruster cells, communication
cells, and solar panel cells (see Figure 2).
The most attractive feature ofcellular satellites is the onorbit
self-reconfiguration capability, which means one cellular
satellite can gradually change its morphology by cell-bycell
moves, with the help ofspace assembly robots (as shown
in Figures 1 and 2). The space assembly robots are responsible
for moving or manipulating the unit cells, and can flexibly
move themselves with/without payload unit cells on free
interfaces by alternating their " head " and " foot, " similar to
crawling robots [11]. The self-reconfiguration ofcellular satellite
is similar to that ofgroundmodular robots [14], such as
the M-tran robot [15] and UBot [16]. However, the selfreconfigurations
ofmodular robots and cellular satellites are
quite different. In modular robots, every module typically
has strong sensing and locomotion abilities. But, in cellular
satellites, most unit cells are simple with single functions;
IEEE A&E SYSTEMS MAGAZINE
JUNE 2022
https://orcid.org/0000-0001-9022-6369 https://orcid.org/0000-0002-7300-3144 https://orcid.org/0000-0002-5132-9602
IEEE - Aerospace and Electronic Systems - June 2022

Table of Contents for the Digital Edition of IEEE - Aerospace and Electronic Systems - June 2022
