69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.

Copyright ©2018 by the International Astronautical Federation (IAF). All rights reserved.

IAC-18,B6,3,7,x43656

Dance: a Frictionless 5 DOF Facility For GNC Proximity Maneuvering Experimental Testing And
Validation

Pierluigi Viscontia*, Stefano Silvestrinia, Michèle Lavagnaa

a
Department of Aerospace Engineering, Politecnico di Milano, Via La Masa, 34, 20156 Milan, Italy,
[email protected]
* Corresponding Author

Abstract
In the last decade, the different space agencies, such as NASA and ESA, have shown serious interest in considering
new mission scenarios exploiting a swarm of satellites flying in formation (i.e., Formation Flying). Simultaneously,
the growing trend of miniaturization perfectly fits this perspective, providing even more flexibility to the mission.
Having a swarm of small spacecrafts, distributed over a large volume, can be used to achieve various capabilities and
functions. It can be exploited for a three-dimensional mapping of the space, providing much more information with
respect to the large and singular satellite. Moreover, the risk at the launch site can be mitigated using different launchers
and launch dates. Even the robustness of the entire mission would be intensely incremented, spreading the redundancies
across the formation rather than across the same monolithic spacecraft. Finally, the overall mission cost can be reduced
exploiting a mass-production technique for the different satellites. Therefore, Formation Flying introduces new
possibilities for the space engineering, but it obviously also poses new challenges due to the low level of maturity of
such technologies. Here comes the need of new test campaigns able to rise the Technology Readiness Level (TRL) of
this innovative hardware and software.
In this context, this paper presents the design, the modelling and the setup of a new frictionless facility at the Aerospace
Science and Technology Department of Politecnico di Milano for on-ground testing and validation of spacecrafts
relative GNC maneuvering. It will serve as a testbed for experimental verification of innovative software, such as
artificial intelligence in the control logic, as well as testing new hardware-in-the-loop. The force-less and torque-less
environment is achieved using a set of linear and hemispherical air bearings. In this way, each vehicle has five degrees
of freedom (DOFs): two translations and three rotations. This also follows the trend of the state of the art, being it a
common architecture adopted for on-ground spacecraft simulators.

Keywords: GNC, on-ground testbed, spacecraft, proximity, frictionless

Nomenclature 𝛒?@ 50 Attitude Platform Centre of Mass Position

𝒈 Gravitational Acceleration
𝑞" Generalized Coordinate 𝒇CD Cold Gas Thruster Force
𝑄$ Generalized Non-Conservative Force 𝒍CD Cold gas Thruster Position
"
𝐿 Lagrangian ⊙ Element-wise Dot Product
𝑇 Kinetic Energy ⊗ Column-wise Cross Product
𝑉 Potential Energy 𝑨H/J Direction Cosine Matrix
𝑋) , 𝑌) , 𝑍) DANCER Position 𝑭/ CGT Total Force Vector
φ, θ, ψ Attitude Platform Euler Angles 𝑴/ CGT Total Torque Vector
ψ/0 Translational Platform Angle 𝑩J CGT Direction Matrix in Inertial Frame
θ12" Reaction Wheel Angle 𝑩H CGT Direction Matrix in Body Frame
𝑋̇) , 𝑌)̇ , 𝑍̇) DANCER Velocities 𝑳 CGT Configuration Matrix
𝝎50 Attitude Platform Angular Velocities 𝒖@ Desired Controlling Torques
𝜔 /0 Translational Platform Angular Velocity ˆ Hat Operator
𝝎12 Reaction Wheels Angular Velocities 𝑨O Attitude Error Matrix
𝑀) DANCER Total Mass 𝑨P Desired Attitude Matrix
𝑀/0 Translational Platform Mass 𝒆5 Attitude Error Vector
𝑀50 Attitude Platform Mass 𝝎O Angular Velocities Errors
𝐽/0 Translational Platform Moment of Inertia 𝐾S , 𝐾5 Attitude Controller Gains
𝑱50 Attitude Platform Moment of Inertia 𝒖U Desired Controlling Forces
𝑨12 Reaction Wheels Configuration 𝒆V Position Error
𝑰12 Reaction Wheels Inertia Matrix 𝒆W Velocity Error
𝛒𝑪𝑴 /0 Translational Platform Centre of Mass Position 𝐾0 ,𝐾X ,𝐾) Translational Controller Gains

IAC-18,B6,3,7,x43656 Page 1 of 11
 69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright ©2018 by the International Astronautical Federation (IAF). All rights reserved.

Acronyms/Abbreviations Flying, define formation flying as "a set of more than one
satellite whose dynamic states are coupled through a
DOF Degree Of Freedom common control law. In particular, at least one member
TRL Technology Readiness Level of the set must (1) track a desired state relative to another
DAER Aerospace Science and Technology member, and (2) the tracking control law must make use
Departmet
of the state of at least one of other members" [2]. This
FCT Formation Control Testbed
JPL Jet Propulsion Laboratory
means that, despite being both distributed systems,
TEAMS Test Environment for Applications of constellations and formation flying satellites completely
Multiple Spacecraft differ in the control strategy. The former consist of a
DANCE Device for Autonomous guidance Navigation number of satellites, each one controlled uniquely with
& Control Experiments respect to its position and velocity. On the contrary,
DANCERS Device for Autonomous guidance Navigation formation flying satellites are controlled in order to
& Control Experiments on Relatively moving maintain a relative position, separation or orientation. For
Spacecrafts these reasons, having a swarm of spacecrafts distributed
IR Infrared over a large volume can provide many advantages: the
TP Translational Platform
AP Attitude Platform
three-dimensional mapping of the space, the reduction of
CM Centre of Mass the risk at the launch site, the reduction of the overall
CR Centre of Rotations mission cost and even the increase in mission robustness.
BSP Board Support Package For all these benefits and many more, the different
OBC On-Board Computer space agencies, such as NASA and ESA, are
SBC Single Board Computer continuously exploiting this architecture in many
RTOS Real-Time operative System different new mission proposals. Moreover, the always
CFRP Carbon Fibre Reinforced Polymer stronger trend of miniaturization perfectly fits the
FEA Finite Element Analysis formation flying scenario, providing even more
PWPF Pulse Width Pulse Frequency
VRT Virtual Reality Toolbox
flexibility to the mission. A recent review of existing
PIL Processor-In-the-Loop formation flying and constellation missions using
nanosatellites [3], counts almost forty missions that
employ multi-satellites with mass smaller than 10 kg.
Just like in every engineering field, also in space
1. Introduction engineering it is very important to test and validate new
On October 4, 1957 the Soviet Union successfully technologies before introducing them in the real
launched Sputnik-1. After almost 4 months, on January environment. However, in-orbit testing can be very
31, 1958 the United States launched Explorer-1. If the expensive and time consuming. For this reason, it is
former marked the start of the Space Age, the latter important to be able to simulate the space scenario on
marked the start of the Space Race [1], fostering the use ground. In this context, this dissertation aims to present
of innovative technologies in the space engineering. Soon the design, the modelling and the setup of a frictionless
arose the need of larger and bigger spacecrafts, as well as facility at the Aerospace Science and Technology
automatic control, in order to fulfil new mission Department (DAER) of Politecnico di Milano for on-
scenarios. However, in the last decade, the various space ground testing and validation of spacecraft relative GNC
agencies have shown serious interest considering new maneuvering. It will serve as a testbed for experimental
mission scenarios exploiting a swarm of small satellites verification of innovative software, such as artificial
flying in formation (i.e., Formation Flying). This intelligence in the control logic, as well as testing some
introduces new possibilities for the space engineering, new hardware-in-the-loop.
but it obviously also poses new challenges due to the low
level of maturity of such technology. Here comes the 1.2 Spacecraft Simulators Architectures
need of new test campaigns able to rise the Technology Since the beginning of the space exploration, the need
Readiness Level (TRL) of this new hardware and of simulating the space environment on ground was
software. evident. No hardware can be launched without a series of
tests and verification. And the same holds for any part of
1.1 Formation Flying in Space a satellite software. However, every component of a
The formation flying satellites are going to spacecraft works in a specific way, meaning that many
revolutionize the space engineering. In fact, the use of a different simulators must be employed. This is probably
cluster of satellites can bring many different advantages, the main reason for the well-known slowness in the
from the launch up to the disposal. But, first of all, it is technological development of the space industry.
important to delineate what formation flying means. In order to recreate the space environment on ground,
Wang, Wu, and Poh, in the book Satellite Formation it is important to simulate the condition of microgravity.

IAC-18,B6,3,7,x43656 Page 2 of 11
 69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright ©2018 by the International Astronautical Federation (IAF). All rights reserved.

Only in this way, it would be possible to test attitude and Moving to Europe, it’s worth mentioning the Test
translational dynamics in a very representative and Environment for Applications of Multiple Spacecraft
effective way. For this reason, several studies have been (TEAMS). It is a test facility developed and built up at
carried out in developing systems able to simulate the Institute of Space Systems of the DLR in Bremen,
weightlessness. Possible solutions for this problem are Germany [6]. It consists of two vehicles, called
the drop towers and aircraft parabolic flights. However, TEAMS_5D, with 5 DOFs and four smaller vehicles,
they can provide only few seconds of simulations. Also called TEAMS_3D, with only 3 DOFs (i.e. planar
underwater test tank can compensate gravity by motion). If the latter are based only on linear air bearings,
buoyancy forces, even though it can be very challenging the former employ configuration with two platforms
to submerge the satellite in water. Another possibility is connected by a spherical air bearing. The TEAMS_5Ds
to create a force-free and torque-free environment where are fully controllable in attitude as well as in translations,
the spacecraft dynamics can be unperturbed. Many in fact, they are equipped with sixteen cold gas thrusters
solutions can be adopted such as magnetic levitation or of 0.06 N and three reaction wheels. The total mass of the
gravity offloading systems. However, the most adopted vehicle is smaller than 100 kg, being therefore smaller
solution for recreating a space-like environment on with respect to the previous facility. However, also the
ground is thanks to air bearings. test arena is smaller, in fact, it is made of two granite
An air bearing is a mechanical device able to create a tables of 4 m-by-2.5 m each. Figure 2 shows the TEAMS
thin film of air between two surfaces in relative motion. facility with the different vehicles.
Therefore, the pressurized air injected in the system acts
as a lubricant and avoids contact between the two
surfaces, reducing the friction of several orders of
magnitude. This architecture for an on-ground spacecraft
simulator has been patented and implemented for the first
time in the 1960 by NASA [4]. However, because of its
several advantages against all the other systems, several
other universities and industries soon adopted this
configuration as well. Moreover, this system can be
adopted both for orbital dynamics simulations as well as
for attitude dynamics simulations. In fact, using a planar
system with linear air bearings, it can provide a force-free
environment for planar motion. On the contrary, using a
rotational air bearing it is possible to create a torque-free
environment. The use in conjunction of linear and
rotational bearings provide a 5-DOF environment Fig. 1 The Formation Control Testbed - JPL [5]
suitable for relative GNC testing and maneuvering.

1.3 Existing Facility Survey

One of the most import on-ground spacecrafts
simulators is the Formation Control Testbed (FCT) at the
Jet Propulsion Laboratory (JPL). It is a multi-robot,
flight-like, system-level testbed for ground validation of
formation GNC architectures and algorithms [5]. Figure
1 shows the selected architecture for the vehicles: a
combined system made of two main platforms. Each
robot is 1.6 m tall and has a diameter of 1.5 m. The lower
platform weights about 80 kg and it hosts three linear air Fig. 2 The Test Environment for Applications of
bearings for translational motion, a spherical air bearing Multiple Spacecraft - DLR Bremen, Germany [6]
for rotational motion and a linear actuator for the vertical
translation. Four composite vessels are employed on the
lower platform for the air storage. The upper platform, on 2. Facility Design
the contrary, weights 170 kg and it exploits sixteen 1 N Formation flying is going to innovate completely the
cold gas thrusters and three orthogonally-mounted space industry, however, it will cross the chasm only
reaction wheels for the dynamics control. The facility when the TRL of the relative technologies, both hardware
floor is made of fourteen metal panels for a total area of and software, will be raised to the maturity level. This
7.3 m-by-8.5 m. implies the testing and the verification of such
technologies. But the only way to cut the costs and

IAC-18,B6,3,7,x43656 Page 3 of 11
 69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright ©2018 by the International Astronautical Federation (IAF). All rights reserved.

shorten the process duration is thanks to on-ground 2.2.1 Attitude Platform

facilities. For this reason, this project wants to provide In order to perform relative dynamical tests, it is
the Politecnico di Milano with a cutting-edge testbed for important to have full controllability of DANCERS
experimental tests and simulations of relative attitude and attitude and translational dynamics. Therefore, the upper
orbital dynamics. platform hosts a set of 12 cold gas thrusters able to
A frictionless facility already exists at DAER provide a thrust of 1.1 N each, modulated at a very high
laboratories, however, it is affected by few flaws. For frequency (1000Hz). The nozzles are placed in such a
this reason, the aim of this paper is to design a completely way so to roughly control both translations and rotations
new simulator able to demonstrate both attitude and of the AP. Moreover, two vessels of 2 L each are
orbital dynamics for internal as well as external projects. employed for the on-board storage of pressurized air at
The name of this new facility is DANCE: Device for 200 bar, ensuring 30 minutes of simulations.
Autonomous guidance Navigation & Control The fine control of the attitude dynamics is in charge
Experiments. Analogously, the related vehicles will be of a set of three reaction wheels, orthogonally-mounted,
called DANCERS: Device for Autonomous guidance able to provide a maximum torque of 55 mNm. All the
Navigation & Control Experiments on Relatively actuation system is custom-made at DAER laboratories.
moving Spacecrafts. It follows a brief description of the However, having the air bearings and a system of
testbed arena as well as of the vehicles. actuators does not ensure the capability of simulating the
microgravity environment. In fact, any static or dynamic
2.1 DANCE Test Environment unbalance of the AP would imply a relevant perturbing
Following the work done by Carlini and Marcuccio in torque on the system. The only way to reduce as much as
[7], the idea is to employ a glass plate, 10 mm thick and possible these torques is by placing the centre of mass
3 m-by-3 m large, as testbed floor for DANCE facility. (CM) of the upper platform as close as possible to the
Even if the crystal has a very high elastic modulus, an centre of rotations (CR) of the hemispherical bearing. On
uniform supporting structure made of self-levelling the other hand, asking for a maximum drift of 5° in one
epoxy resin is employed to ensure system rigidity. minute, it implies a maximum error between CM and CR
Moreover, in order not to introduce relevant perturbing of 10-7 m, and this is not achievable merely by a well-
forces to the vehicles, the maximum allowable floor designed configuration.
slope is computed to be only 0.38 mrad. Therefore, a complex system of masses is used as
Another important aspect is the facility tracking balancing system for the entire platform. Moreover, the
system, able to provide an absolute measurement of the potential on-board payload could introduce dynamical
vehicles position and attitude. It has to meet very strict unbalance. Thus, the balancing system consists of a static
requirements in terms of accuracy. In fact, the selected balancing system (Fig. 3-a) for gross and manual
system employs a set of four infrared (IR) cameras and corrections and a dynamic balancing system (Fig. 3-b)
different markers on the vehicles: in this way it is able to for automated and precise offset compensation during the
provide DANCERS position within few millimetres and entire simulation. Adopting a system with three
their attitude within 0.4◦. orthogonally-mounted lead-screws with a very small
Finally, an air filling system is employed to refill the pitch, it is possible to finely move three masses so to fulfil
vessels for the vehicles on-board air storage and a control the balancing requirements.
room provides the user with all the different tools and DANCER vehicles have to perform dynamics control
supporting hardware in order to conduct and monitor the autonomously, employing the on-board hardware. This
entire simulation. implies that all the sensors, especially those concerning
attitude dynamics and translational dynamics must be
2.2 DANCERS able to provide a state estimation without relying on
In order to test and validate relative GNC external systems. Therefore, an Inertial Measurement
maneuvering, the facility employs two vehicles able to Unit (IMU) is employed together with a compass sensor,
recreate a torque-free and force-free environment in 5 in order to have absolute reference knowledge.
DOFs. At this purpose, each vehicle mounts three linear Moreover, an optical camera is used for relative state
air bearings for the planar motion and a hemispherical air estimation between the two vehicles.
bearing for the rotational motion. Thus, DANCER can be For what concerns the avionics, the key points for the
seen as made of two distinct platforms: the Translational selection are: cost, mass, volume, performances and
Platform (TP) on the bottom of the robot and the Attitude availability of the board support packages (BSP) for
Platform (AP) on top. The former has to provide the Matlab® and Simulink® environment. This last point is
frictionless environment for the latter, which, on the important because, nowadays, the MathWorks® Suite
contrary, can be seen as the actual spacecraft able to host has become a very powerful tool for engineering and it is
a possible payload. important to provide the facility of complete Matlab and
Simulink support. In this way, it will be possible, thanks

IAC-18,B6,3,7,x43656 Page 4 of 11
 69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright ©2018 by the International Astronautical Federation (IAF). All rights reserved.

to the embedded coder, to easily convert Simulink blocks frictionless capabilities. In fact, it shall be able to
into equivalent code to be quickly uploaded on the reproduce a force-free and torque-free environment in the
vehicle on-board computer (OBC). Accordingly, the planar motion as well as in the three attitude rotations.
baseline architecture uses an ARM®-based single board The solution adopts three linear air bearings and a
computer (SBC) as main OBC working side by side with hemispherical air bearing. The former are flat pads
a microcontroller from Texas Instrument. On one hand, equipped with small holes where pressurized air flows,
the SBC will have a real-time operative system (RTOS) creating an air cushion on which the system translates.
able to run high-level algorithms such as control laws, They are selected having in mind the total mass of
optical navigation, artificial intelligence and other DANCER vehicle. On the contrary, the latter consists of
complicated software. On the other hand, the a pair of concave-convex hemispheres perfectly lapped.
microcontroller will serve as intermediary between OBC The concave base is made by a porous material where air
and hardware. flows, creating the air cushion. It is selected considering
its load capacity since it has to withstand the weight of
the attitude platform. Moreover, a pneumatic circuit is
employed in order to provide air to the floating system
for the entire duration of the simulation. At this purpose,
two carbon-fibre vessels of 3 L each are used to store on
board 1200 L of dry air.
The last component of DANCERS, and perhaps the
most important, is the structure, which has to
accommodate all the aforementioned components in an
(a) (b) optimal configuration. Having in mind the main driver
that is rigidity, the structure is designed starting from its
Fig. 3 DANCERS Static Balancing System (a) and main elements. In fact, the key elements to be considered
Dynamic Balancing System (b) are:
- it has to mount the three linear air pads on the very
The communication between the two DANCERS and bottom, offloading all the vehicle weight on them
the control room is performed using an RF link, in order and keeping the CM at the centre of the contact
to resemble as much as possible a real intersatellite-link. points;
Thus, a 100mW RF transceiver at 433 MHz is used to - it has to mount the hemispherical air bearing to
transfer more than 50 state variable, 2 bytes each, at a enable free rotations of the system, therefore, a
frequency of 100 Hz. pedestal structure is needed in order to avoid any
An electric power system has to provide power to all possible obstruction of the free-floating platform
the actuators and the avionics on board. However, the movements;
vehicles cannot have any mechanical interface with the - the free-floating platform has to host all the different
external world because it would perturb its dynamics. For components, trying to keep the CM and the CR as
this reason, a system of batteries is used to store energy close as possible.
on board. Considering the power budget of the vehicle Here comes the need of two distinct platforms: the TP
(Tab. 1), the battery pack uses 16 cells of 3,7 V each, able to host the floating system and the AP where all the
arranged in such a way so to obtain a capacity of 6000 other subsystems are placed. Moreover, to align as much
mAh, a sufficient amount of energy to cover the entire as possible CM with CR, AP can be subdivided into two
duration of the simulation. other parts. Placing the massive actuators below the
centre of rotations and the avionics, with the electronics
Subsystem Power [W] (20% margined) and the payload, above it, it is possible to balance the
Actuators 217.2 entire system.
Balancing System 16.8 All the structure will be made in-house using the
Sensors 5.9 carbon fiber reinforced polymer (CFRP). In this way it is
Avionics 23.0 possible to save mass, gaining in strength and rigidity. In
Total 262.9 fact, the vehicle structure has to withstand the weight of
Total + 10% 289.2 all the subsystems, trying to minimize the deflections due
to system dynamics. Considering DANCERS mass
Tab. 1 DANCER Power Budget budget (Tab. 2) and the relative configuration of the
different subsystems (Fig. 4), the structure sizing has
2.2.2 Translational Platform been performed using a Finite Element Analysis (FEA)
One of the most important parts of DANCERS in order to assess the stress conditions of the different
vehicles is the floating system, which will ensure the structural components.

IAC-18,B6,3,7,x43656 Page 5 of 11
 69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright ©2018 by the International Astronautical Federation (IAF). All rights reserved.

- CM and CR of the attitude platform precisely

Subsystem Mass [W] (20% margined) coincide for the entire duration of the simulation;
(1) Floating System 9.4 - No friction is present;
(2) Actuators 10.9 - The testbed arena is exactly flat and horizontal.
(3) Balancing System 9.9 The set of differential equations describing the attitude
(4) Sensors 0.3 dynamics and the translational dynamics of the vehicle
(5) Avionics 0.1 can be derived exploiting the Euler-Lagrange equations.
(6) Electric Power System 3.1 The generic form for a set of generalized coordinates 𝑞" ,
(7) Structure 5.9 in the presence of non-conservative forces 𝑄$ , is:
(8) Payload 2 "

Total 41.6
P [\ [\
Total + 20% 49.9
PY
([$̇ ) − [$ = 𝑄$a ∀ 𝑖 = 1, . . . , 𝑛 (1)
] ]

Tab. 2 DANCER Mass Budget where L is the Lagrangian, made as:

𝐿 = 𝑇−𝑉 (2)

T and V are the kinetic energy and the potential energy of

the system, respectively. On the other hand, the
generalized coordinates 𝑞" can be identified considering
the degrees of freedom of the entire system, which are:
- 𝑋) , 𝑌) and 𝑍) are the translations of the TP and of
the AP;
- φ, θ and ψ are the euler angles for the AP;
Fig. 4 DANCER Configuration - ψ/0 is the rotation of the TP around the vertical axis;
- θ12h , θ12i and θ12j are the angles of the three
reaction wheels.
3. Dynamical Model These coordinates are state variables of the system,
The DANCE facility will serve as a simulator for
together with:
relative dynamics, both translational and rotational.
- 𝑋̇) , 𝑌)̇ and 𝑍̇) are the velocities of the TP and of the
Therefore, dynamical models of the two vehicles on the
AP;
platform will be fundamental. Only in this way, it will be
- 𝝎50 is the vector of the angular velocities of the AP;
possible to predict and control in the right way the
DANCERS dynamics and accomplish the simulation - 𝜔/0 is the angular velocity of the TP around the
goals. vertical axis;
However, before starting with relative dynamics, it is - 𝝎12 is the vector of the angular velocities of the
important to set-up the entire facility. Actuators have to three reaction wheels.
be characterized, sensors have to be calibrated, control The total kinetic energy of the entire robot can be seen as
laws have to be tuned. All the components of the system the sum of the kinetic energies of the different
will have to be tested and validated. For these reasons, components of DANCER, which are considered as rigid
before developing a relative dynamical model, it is bodies. Therefore, under the aforementioned
important to derive the model of the single robot with all assumptions, the kinetic energy can be computed as:
its main components. It will serve for initial numerical h P𝒓 P𝒓 h h
tests, for the tuning of the control laws and also as 𝑇 = i 𝑀) k PYmn ⋅ k PYmn + i 𝐽/0 𝜔/0
i
+ i 𝝎50 ⋅ 𝑱50 𝝎50 +
visualization tool. Then, in a second stage, it will be used h
𝝎12 ⋅ 𝑨12 𝑰12 𝝎12 + 𝝎50 ⋅ 𝑨12 𝑰12 𝝎12 (3)
as fundamental brick for the development of a refined i
relative dynamical model of the facility.
where 𝑀) , 𝐽/0 and 𝑱50 are respectively the total mass,
3.1 Equations of Motion the moment of inertia of the translational platform and
In order to derive the equations of motion describing the inertia matrix of the attitude platform in principal
the vehicle dynamics, it is important to make few axes of inertia. The matrix 𝑨12 is introduced to take care
assumptions: of the reaction wheels orientation and, in this case,
- DANCER is considered as made by discrete rigid having them aligned with the body reference frame, it is
bodies, in particular the translational platform, the equal to the identity matrix. On the contrary, 𝑰12
attitude platform and the reaction wheels; contains, on its diagonal, the moments of inertia of the
three wheels.

IAC-18,B6,3,7,x43656 Page 6 of 11
 69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright ©2018 by the International Astronautical Federation (IAF). All rights reserved.

Having a system of rigid bodies, the only potential These equations can be solved thanks to a numerical
energy is the gravitational energy. It can be computed computation software, starting by a set of initial
considering the vectors 𝛒𝑪𝑴 /0 and 𝛒?@50 describing the conditions and using a controller to provide system
position of the centers of mass with respect to the center stability.
of rotation. Therefore:
3.2 Numerical Model
𝑉 = 𝑀/0 𝒈 ⋅ q𝒓) − 𝝆?@st u + The behaviour of the cold gas thrusters is highly non-
𝑀50 𝒈 ⋅ (𝒓) − 𝝆?@vt ) (4) linear and non-conservative. Consequently, a
computational software has to be employed for the
solution of the problem. With this aim, the Mathworks®
where 𝒈 is the vector of the gravitational acceleration.
suite is exploited, especially Matlab® and Simulink®.
Considering the assumptions, it is possible to
The entire model of DANCER is implemented in the
demonstrate that potential energy is constant.
software, trying to simulate its dynamics. Equations of
For what concern non-conservative forces, no friction
motion, actuators and sensors are introduced as blocks.
terms are involved in DANCERS dynamics because of
The direction cosine matrix is employed to describe
the air bearings. However, they employ a set of cold gas
the vehicle kinematics. In fact, it is the most powerful
thrusters which generates non-conservative forces and
attitude representation because it is global and unique.
torques on the robots. It is possible to write their
Moreover, it is important to introduce a controller able to
contribution considering the four cluster:
command the thrusters and the reaction wheels in order
z to reach a desired dynamic. For what concern the attitude
𝑭/ = wCyh ∑jDyh 𝒇CD (4) dynamics, a non-linear controller is derived using the
Lyapunov’s direct method on a function of the energy of
z
𝑴/ = wCyh(∑jDyh 𝒍CD × 𝒇CD ) (5) the system. It has the following form:

where 𝒇CD is the vector of the force exerted by the n-th 𝒖@ = −𝐾S 𝑱50 𝝎O − 𝐾5 𝒆5 +
thruster on the k-th cluster, while 𝒍CD is the position 𝝎50 × 𝑱50 𝝎0 − 𝑱50 [𝝎O ] 𝑨
̂ O 𝝎P (12)
vector of the n-th thruster on the k-th cluster in body
frame. These equations can be simplified adopting a where the symbol ˆ represent the hat operator used to
matrix notation: obtain a matrix from a column vector, 𝑨O is the error
matrix and it is equal to the identity matrix only when
𝑭/ = 𝑩J 𝒇 / (6) 𝑨H/J is equal to the desired 𝑨P . 𝒆5 is the attitude error
computed from 𝑨O using the inverse hat operator. The
𝑴/ = 𝑳 ⊗ (𝑩H ⊙ 𝒇 / ) (7) vector 𝝎O , instead, contains the errors in the angular
velocities with respect to the desired 𝝎P . Finally, 𝒖@ is
where the operator ⊙ represents the element-wise dot the vector containing the desired torques while 𝐾S and
product while ⊗ represents the column-wise cross 𝐾5 are the two gains used to tune the system response for
product. These two operations are fast and easy to handle what concern the attitude dynamics.
on numerical software such as Matlab. On the other hand, for what concern the translational
Manipulating all the previous equations it is possible to dynamics, a simple PID controller is adopted and the
obtain the equations of motion for the DANCER vehicle: desired forces 𝒖U can be computed as follows:
. Y
𝑱50 𝝎50 + 𝝎50 ×. 𝑱50 𝝎50 = 𝑨12 𝑰12 𝝎12 × 𝒖U = 𝐾0 𝒆V + 𝐾X ∫‰ 𝒆V (𝑡 Š )𝑑𝑡 Š + 𝐾) 𝒆W (13)
𝝎50 − 𝑨12 𝑰12 𝝎12 + 𝑳 ⊗ (𝑩H ⊙ 𝒇 / ) (8)
where 𝒆V is the error with respect to the desired position
P | 𝒓m
𝑀) = 𝑨/H ⁄J 𝑩H 𝒇 / (9) and 𝒆W is the error with respect to the desired velocity.
PY |
The gains 𝐾0 , 𝐾X and 𝐾) can be tuned in order to
where 𝑨H/J is the direction cosine matrix used to rotate optimize the system response.
It is also important to implement in the model the
a vector from inertial reference frame to body reference
behaviour of the actuators. The term in equation 8 related
frame. The constraint equations are obtained as well:
with the reaction wheels represents their torque and it has
to be limited in order to represent the real system
𝜔/0 = 0 → 𝜓/0 = 𝑐𝑜𝑛𝑠𝑡 (10)
capabilities. On the other hand, the cold gas thrusters are
jets of pressurized air controlled by fast electrovalves.
𝑌)̇ = 0 → 𝑌) = 𝑐𝑜𝑛𝑠𝑡 (11) Therefore, there is no way to control the amount of force
they can exert. The only way to achieve some flexibility

IAC-18,B6,3,7,x43656 Page 7 of 11
 69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright ©2018 by the International Astronautical Federation (IAF). All rights reserved.

is adopting a modulation scheme which is able to provide

a sort of throttleability to the system. In this model, a
Pulse-Width-Pulse-Frequency (PWPF) modulation
scheme is adopted because of its benefits with respect to
other algorithms. In fact, it is able to save propellant
mass, providing a smoother actuation with a close-to-
linear behaviour. More information on thrusters PWPF
modulation can be found in the work of Agrawal,
Mcclelland, and Song [8]. Another important aspect of
this subsystem is how to implement the thruster selection
algorithm. Lately, a lot of research has been done in this
field, trying to find the most efficient way to select the
right thrusters in order to reduce the propellant
consumption. However, this is at odds with Fig. 5 VRT DANCER graphical model
computational cost, therefore, a trade-off is necessary.
The baseline is to adopt a jet selection by look-up table.
This is a very fast and easy algorithm, which can be
optimized employing an efficient reference catalogue. 4. Facility Integration: Current Status
Finally, the external constraints have to be applied on The following section describes the current status of
the equations of motion. In fact, the vehicle position and the implementation of the DANCE test-bed. In particular,
attitude are not completely unconstrained. The platform the structure, the propulsion system and pneumatic
has a finite area of 3 m-by-3 m. On the other hand, the management system are being developed and
hemispherical air bearing is not able to provide full manufactured. The Microcontroller tests are focused on
rotations in pitch and roll. This can be achieved assessing the most effective approach for fast prototyping
modelling a damper-spring system on the border of the and translation of control algorithms from numerical
arena as well as at the limits of the bearing. simulations to hardware implementation.
Thanks to all the different aforementioned blocks, the
entire model of the DANCER dynamics is built. The 4.1 Primary Structure
model can be numerically solved starting from a set of In order to save mass in the primary structure, the
initial conditions and adopting the right solver. In fact, it composite carbon fiber material is selected for massive
is important to consider that the resulting system of components. A pre-impregnated carbon-epoxy material
differential equations is stiff because the dynamics of the is utilized for manufacturing at the Department of
vehicle is much slower with respect to the dynamics of Aerospace Science and Technology laboratory (DAER)
the thrusters, for example. Therefore, a fixed-step third- at Politecnico di Milano. In particular, the selected
order method is employed for the numerical integration product is a 300 mm unidirectional tape from HEXCEL.
of the entire model. The resulting numerical model has The lamination sequence is a composition of plies
been exploited for the sizing of the major components of orientation 0°, 90°, -45° and 45°. The laminated plates
each subsystem of the DANCER vehicles. It has also have been cured in the autoclave following the desired
been used for the initial characterization of the system cycle of pressure and temperature. Errore. L'origine
performances. Moreover, it will serve as starting point for riferimento non è stata trovata.Figure 6 and 7 show the
the implementation of the DANCER software. different stages of the manufacturing process: the plies
In order to visualize the complex dynamics of the lamination and the subsequent autoclave curing at DAER.
DANCER vehicle, it is possible to exploit the Virtual
Reality Toolbox (VRT) which allows to create animated
three-dimensional scenes driven from the Matlab and
Simulink environment. A screen-shot is reported in
figure 5 as example.

Fig. 6 Plies lamination

IAC-18,B6,3,7,x43656 Page 8 of 11
 69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright ©2018 by the International Astronautical Federation (IAF). All rights reserved.

The last component of the DANCER primary structure

made by a composite material is the pedestal body,
shown in Errore. L'origine riferimento non è stata
trovata. 10.

Fig. 7 Autoclave curing

The plates have been cut using a water jet machine

shaping the external circumference and the big-sized
holes. Figure 8 shows the base reinforcement plate after
the water jet cut.

Fig. 10 Pedestal

4.2 Propulsion System

The propulsion system is composed of cold-gas thrusters,
operating with compressed air. Given the on-ground
application, it is unnecessary to buy space-qualified
hardware; hence, the thrusters’ assemblies are
manufactured and assembled in-house. In order to obtain
thrust from a compressed fluid in an efficient way, it is
important to adopt a nozzle. A commercial nozzle has
been chosen for the implementation in order to deliver a
nominal thrust of 1.1 N at a nominal pressure of 6 bar.
The single thruster assembly, shown in figure 11Errore.
Fig. 8 Base reinforcement plate after water jet cut. L'origine riferimento non è stata trovata., is composed
by three nozzles mounted on the aluminium housing. The
Due to the risk of delamination, the smaller holes were aluminium housing provides the fluidic connection
drilled using drill bits provided by SANDVIK®. Figure between the nozzles and the pneumatic management
9 shows the Middle Plate with all the holes drilled at system.
DAER.

Fig. 9 Middle Plate after drilling Fig. 11 Thruster assembly

IAC-18,B6,3,7,x43656 Page 9 of 11
 69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright ©2018 by the International Astronautical Federation (IAF). All rights reserved.

The aluminium housing is secured to the structure by four mm fluidic tubing connections and 12 in-line
screws and the pneumatic line interface is made of three fast switching electrovalves (1000 Hz), each
channels connecting the fluidic fittings to the nozzles. one controlling the flow through one single
The complete propulsion system comprises four nozzle.
assemblies, yielding a total of 12 thrusters, placed
orthogonally on the xy-plane of the attitude platform, in
order to provide controllability along every axis, both
rotationally and translationally.

4.3 Pneumatic Management System

The pneumatic management system concerning
pressures below 10 bar completely relies on components
by FESTO®. Two separated assemblies are integrated,
namely one for the Attitude Platform and one for the
Translational Platform. The former is in charge of
delivering compressed air to the linear and the
hemispherical bearings, with a nominal pressure of 5.5
bar; the latter is dedicated to the feeding of the propulsion
system. The microcontroller, which is responsible for the
low-level control architecture, handles part of the
actuation, e.g. electrovalves.
Fig. 13 Pneumatic management system on the Upper
• Translational Platform: the TP pneumatic Platform
management system, shown in figure 12Errore.
L'origine riferimento non è stata trovata., The integration between the pneumatic circuit and the
comprises a proportional pressure regulator, 6- primary structure is currently being performed.
mm fluidic tubing connections and two directly
actuated solenoid valves, together with generic 4.4 MCU/SBC Testing
fluidic fittings. One pneumatic line is dedicated The DANCER on-board computing process is split into
to the hemispherical bearing, whereas a single two hierarchical levels. As mentioned, DANCER is
valve controls the flow through the three linear equipped with a BeagleBone Black Wireless with
bearings. onboard 802.11 b/g/n 2.4GHz WiFi and Bluetooth
mounting a Octavo Systems OSD3358 1GHz ARM®
Cortex-A8 processor as Single Board Computer in
charge of handling high-level task, data processing and
control algorithms.
The low-level controller is a C2000 Delfino MCUs
F28379D, which is in charge of controlling the hardware.
For instance, the MCU unit manages the electrovalves
actuation, RW speed/torque control and dynamic
balancing. The control algorithms are developed in
Matlab/Simulink® environment and transferred into the
boards using the dedicated coder. This approach is
adopted to ease the process of algorithm implementation
for fast prototyping as well as to increase the
standardization for third parties interface.
Figure 14 shows the result of a Processor-in-the-loop
(PIL) simulation in which the DANCE control algorithm
is deployed to the board and run in parallel with a
Fig. 12 Pneumatic management system on the Desktop-PC equipped with Intel® Core™ i5-3470
Lower Platform [email protected] 8GB RAM.

• Attitude Platform: the AP pneumatic As shown in figure 14, the discrepancies are <10-6%,
management system, shown in figure 13Errore. meaning that the board is suitable to process the low-
L'origine riferimento non è stata trovata., level control scheme.
comprises a proportional pressure regulator, 6-

IAC-18,B6,3,7,x43656 Page 10 of 11
 69th International Astronautical Congress (IAC), Bremen, Germany, 1-5 October 2018.
Copyright ©2018 by the International Astronautical Federation (IAF). All rights reserved.

provide similar performances in term of experiment

duration and accuracy.

In order to obtain a fully functional facility some actions

need to be taken in the immediate future, mostly
regarding the manufacturing and integration of the
facility itself.

• Functional tests of all the subsystems

• Integration of the secondary subsystems on the
primary structure
• Integration of DANCE test environment: the
resin basement for the arena floor has to be
realized to fix alignment issues.
• Tracking system installation: for absolute
navigation of DANCER.

Fig. 14 Processor-in-the-loop (PIL) simulation of the

control algorithm. The code is deployed to the board Acknowledgements
after translation from Simulink model. The authors would like to thank FESTO, SANDVIK for
the support to the project.

5. Conclusions and Future Steps References

The design and integration of a frictionless 5-DOF
facility for GNC proximity manoeuvring experimental [1] Macdonald, Malcolm and Badescu, Viorel, The
testing and validation have been thoroughly presented in international handbook of space technology, Springer,
this paper. The DANCE facility will be an essential tool 2014.
for testing and validating a variety of possible hardware [2] Wang, Danwei, Wu, Baolin, and Poh, Satellite
and software. Different configurations will be available, Formation Flying. Vol. 87. Jan. 2017.
involving relative dynamics experiments but also [3] Bandyopadhyay, Saptarshi et al, Review of
absolute dynamics in force-free and torque-free Formation Flying and Constellation Missions Using
environment. Formation flying, in fact, is only one Nanosatellites, In: 53 (Mar. 2016), pp. 1–12.
possible scenario that DANCE platform will be able to [4] L, H.A. and W. Haeussermann, Satellite motion
provide. In this way, DAER will benefits of a competitive simulator, US Patent 3,092,918. 1963,
and outstanding tool. Starting from the overall https://www.google.com/patents/US3092918.
requirements and goals, the different subsystems of the [5] Regehr, M. W. et al, The Formation Control Testbed.
DANCERS vehicles, as well as of the testbed arena, have In: 2004 IEEE Aerospace Conference Proceedings
been designed, trading-off various aspects. Several (IEEE Cat. No.04TH8720). Vol. 1. 2004, 564 Vol.1.
difficulties have been faced during this process, because [6] Daitx, Henrique et al. Development of a Combined
of the complex systems involved and because of the Attitude and Position Controller for a Satellite Simulator.
several interconnections within the distinct components. Sept. 2016.
However, in order to deal with the aforementioned [7] Carlini, Alessandro and Marcuccio, Mirna. Apparato
problems, a series of numerical tests have been sperimentale per la simulazione dinamica di un mini-
performed, employing the derived dynamical model. robot flottante con applicazione spaziale. Anno
Indeed, despite necessary assumptions, the DANCER Accademico 2002-2003. MA thesis. Politecnico di
model can be accurate enough for the characterization of Milano.
several aspects of the vehicles. [8] Agrawal, Brij N, Mcclelland, RS, and Song,
After all the iterations and the trade-offs, the resulting Gangbing. Attitude control of flexible spacecraft using
vehicles are small and light, but they still provide all the pulse-width pulse-frequency modulated thrusters. In:
necessary capabilities required for the facility. In fact, Space Technology- Kedlington 17.1 (1997), pp. 15–34.
comparing DANCERS with other 5 DOFs vehicles, the
former are five time lighter than the latter, but they

IAC-18,B6,3,7,x43656 Page 11 of 11