ORION – Massachusetts Institute of Technology

 

 

ORION: A LOW-COST DEMONSTRATION OF

FORMATION FLYING IN SPACE USING GPS

AIAA-98-4398

Jonathan P. How

, Robert Twiggs

, David Weidow

,

(cid:3)

y

z

Kathy Hartman

, and Frank Bauer

x

{

Abstract

operations costs through enhanced vehicle auton-

This paper describes research on the design of a GPS

omy.

based relative position and attitude sensing mission

Recent results have demonstrated that Carrier-

called Orion. The current Orion design consists of

Phase Di(cid:11)erential GPS (CDGPS) techniques can be

a (cid:13)eet of 6 micro-satellites launched simultaneously

used to autonomously track and then control the

into low Earth orbit to demonstrate coordinated at-

relative position and attitude between several space-

titude control, relative navigation and control, and

craft [1, 2, 3, 4, 5, 6, 7, 8]. This sensing technology

formation initialization techniques in space. This

can be used to develop a virtual spacecraft bus us-

approach represents a new systems architecture that

ing automatic control of a cluster of micro-satellites

provides many performance and operations advan-

to replace the monolithic bus of current Earth Sci-

tages, such as reduced operating cost, enhanced mis-

ences Enterprise (ESE) satellites (such as Landsat-

sion (cid:13)exibility, and improved science observations.

7) [3, 6]. Many future space applications would ben-

A key ob jective of Orion is to demonstrate Carrier-

e(cid:12)t from using this formation (cid:13)ying technology to

Phase Di(cid:11)erential GPS (CDGPS) techniques to au-

perform distributed observations, including: earth

tonomously track the relative position and attitude

mapping (SAR, magnetosphere), astrophysics (stel-

between several spacecraft. Based on a research pro-

lar interferometry), and surveillance.

gram focused on low-cost spacecraft design, Orion

The goal is to accomplish these science tasks us-

represents a critical step towards the realization of

ing a distributed array of much simpler, but highly

formation (cid:13)ying and virtual platform capabilities.

coordinated, vehicles (e.g., micro-satellites). This

1

1

Introduction and Motivation

provides many performance and operations advan-

approach represents a new systems architecture that

A revolution in spacecraft guidance, navigation and

tages, such as:

control technology has started through the use of

1. Enables extensive co-observing programs to be

GPS to autonomously provide vehicle position, atti-

conducted autonomously without using exten-

tude, and timing information. Not only will these

sive ground support, which should greatly re-

innovations result in signi(cid:12)cant reductions in the

duce operations cost of future science missions.

weight, power consumption, and cost of future space-

craft attitude and orbit determination systems, they

2. Increased separation (baseline) between instru-

should also result in a signi(cid:12)cant reduction in ground

ments could provide orders of magnitude im-

provement in space-based interferometry. A

AIAA member, Dept. of Aeronautics and Astronautics,

distributed array of spacecraft will signi(cid:12)cantly

Stanford University, Durand Building, Room 277.

AIAA Member, Stanford University

improve the world coverage for remote sensing,

AIAA Member, NASA GSFC

and will enable simultaneous observations us-

AIAA Member, NASA GSFC

AIAA Senior Member, NASA GSFC

ing multiple di(cid:11)erent sensors.

(cid:3)

y

z

x

{

1

Copyright

1998 by the American Institute of

c
(cid:13)

3. Replacing the large complex spacecraft of tra-

Aeronautics and Astronautics, Inc. All rights reserved.

ditional multi-instrument observatories with an

array of simpler micro-satellites provides a (cid:13)ex-

American Institute of Aeronautics and Astronautics

1

ible architecture that o(cid:11)ers a high degree of

within a fraction of a wavelength of light. To achieve

redundancy and recon(cid:12)gurability in the event

this, a layered control approach has been proposed,

of a single vehicle failure.

one layer of which will use CDGPS type sensing to

4. Places the design emphasis on building and (cid:13)y-

regulate the formation positions to within a cen-

ing the science instruments, not on the devel-

timeter [8]. Very precise formation (cid:13)ying of this

type would also be required for distributed SAR and

opment of the bus platform itself. Allows stan-

dardization of the satellite fabrication process,

Earth imaging missions.

which will reduce costs.

The second is the EO-1 mission which is planned

5. Enables the low cost, short lead-time instru-

to be a co-(cid:13)yer with the Landsat 7 spacecraft. The

ments to be built, launched, and operated im-

mediately. The more costly, long lead-time in-

struments can then join the (cid:13)eet when avail-

able. This approach would also allow new or

replacement instruments to join the formation

as they are developed or needed.

scienti(cid:12)c goal of the EO-1 mission is to validate the

results obtained with the multi-spectral imager on-

board Landsat 7 by taking images with a similar in-

strument onboard the EO-1 spacecraft. To achieve

this, the two spacecraft must be (cid:13)own in formation

so that the relative distance between them can be

controlled such that the EO-1 imager is viewing the

Earth through the same column of air as the Land-

curacy on the order of 10-20 m, which is an example

of coarse formation (cid:13)ying [1]. The initial ob jective

was to demonstrate formation (cid:13)ying using the EO-1

However, these bene(cid:12)ts come at a cost because

sat 7 imager. This will require a formation (cid:13)ying ac-

the new systems architecture poses very stringent

challenges in the areas of:

1. Onboard sensing required to perform the au-

and Landsat 7 spacecraft, but because of a variety

tonomous closed-loop relative navigation and

of budget and time constraints, no cross link will

attitude determination and control.

be possible between the two spacecraft. EO-1 will

2. High-level mission management to enable task

demonstrate onboard closed-loop autonomous orbit

allocation across the (cid:13)eet of spacecraft.

control using AutoCon

, which represents a key

TM

3. High-level fault detection recovery to enhance

the mission robustness.

stration of formation (cid:13)ying spacecraft.

step. However this will not provide a true demon-

As a result, the Orion mission was developed

As these points suggest, the onboard autonomy of

to demonstrate the true formation (cid:13)ying concept

these spacecraft must be signi(cid:12)cantly enhanced to

that involves several spacecraft (6) navigating col-

reduce ground support required. Several research

lectively. This mission will be used to validate key

e(cid:11)orts are focused on this problem [9, 10]. Because

sensing and control issues associated with formation

these tools can be extended to the formation (cid:13)ying

(cid:13)ying, and it represents an important step towards

problem, we will not discuss this issue further.

the virtual platforms envisioned for future Earth Sci-

Strong interest in the formation (cid:13)ying concept

cent work on Orion with a particular focus on the

ences Enterprise missions. This paper discusses re-

has developed as a result of two missions that are

GPS sensing, the mission plan, and the spacecraft

currently under development as part of the NASA

bus design. The micro-satellite design is based on a

New Millennium Program (NMP). The (cid:12)rst, the New

modi(cid:12)ed version of the low cost, low weight space-

Millennium Interferometer (NMI), is a formation of

craft bus developed by Stanford University and the

three spacecraft in either GEO or solar orbit (0.1

amateur radio satellite community called SQUIRT

AU from the Earth) to be used for long baseline op-

(Satellite Quick Research Testbed) [12].

tical stellar interferometry [11]. Two of the space-

craft will be light collectors”, separated by several

kilometers, that focus light from a distant star onto

2 Formation Flying Testbeds

a third combiner” spacecraft that forms the inter-

Precise formation (cid:13)ying requires an accurate mea-

ference pattern. To form this pattern, the optical

surement of the formation states, i.e., the relative

path between the spacecraft must be controlled to

American Institute of Aeronautics and Astronautics

2

One of the limitations of GPS for ground test-

ing these formation (cid:13)ying concepts is the reliance

on observability to the twenty four NAVSTAR satel-

lites in Earth orbit. This precludes the use of GPS

indoors. This limitation has been overcome by the

use of pseudolites: transmitters that emit GPS-like

signals [16]. Pseudolite technology enables the ap-

plication of GPS techniques to situations where the

NAVSTAR spacecraft are not visible (e.g.

indoors

or deep space). Pseudolites can also be used to aug-

ment the existing NAVSTAR system as well. Recent

Fig. 1

: Formation Flying Testbed

results have demonstrated the feasibility of using a

attitude and positions of the vehicles. GPS provides

achieve relative navigation accuracies on the order of

a promising technique for sensing these variables at

(cid:14)

2 cm and 0.5

on this formation of prototype space

a much lower cost than combinations of conventional

vehicles. Refs [6, 17] provide more details on this

CDGPS based sensing system (with pseudolites) to

spacecraft sensors such as star trackers, horizon/sun

experimental facility.

sensors, and inertial measurement systems. These

techniques are based on Carrier-Phase Di(cid:11)erential

A second testbed has recently been developed to

GPS (CDGPS) that was developed to improve the

demonstrate formation (cid:13)ight in three dimensions us-

accuracy of GPS measurements [13]. CDGPS is a

ing lighter-than-air vehicles (blimps) [18]. The blimp

relative position measurement technique that helps

formation is also operated indoors, but in a large

circumvent S/A and many other error sources in the

highbay. This second testbed will be used to demon-

basic GPS measurement. Given GPS measurements

strate that various GPS errors, such as the circular

at two nearby antennas, relative position between

polarization e(cid:11)ect, can be modeled and eliminated

these antennas can be estimated to a high degree

from the measurement equations. The errors were

of accuracy based on tracking the relative phase of

not present in the 2D testbed because the antennae

the GPS carrier waves. Of course, there is an in-

on each vehicle all roughly point in the same direc-

teger number of wavelengths di(cid:11)erence between the

tion, but the errors will play a crucial role on-orbit

phases measured at the two antennas. This inte-

because the spacecraft can undergo more general 3D

ger ambiguity is not directly observable and must

motions.

be estimated or calibrated by an additional sensor.

These testbeds are being used as a precursor to

Nevertheless, CDGPS has been used successfully in

the Orion (cid:13)ight vehicles to validate the GPS sensing

a number of applications including automatic preci-

algorithms.

In the future, they will also be used

sion landing of commercial aircraft [14], and attitude

to test the (cid:13)ight control software, the inter-vehicle

estimation for spacecraft in Earth orbit [15].

communication, and the actuators.

To investigate the guidance, navigation, and con-

trol issues associated with precise formation (cid:13)ying,

3 GPS Estimation Issues

a formation (cid:13)ying testbed has been created in the

Stanford Aerospace Robotics Laboratory [2, 3, 6,

This section provides a brief overview of the equa-

17]. The testbed consists of 3 active free-(cid:13)ying ve-

tions used to estimate the relative position and at-

hicles that move on a 12 ft (cid:2) 9 ft granite table top

titude of the vehicles using the GPS carrier phase

(see Figure 1). These air cushion vehicles simulate

measurements. We (cid:12)rst consider the approach used

the zero-g dynamics of a spacecraft formation in a

for the ground testbeds an then discuss the changes

horizontal plane. The vehicles are propelled by com-

required to use this measurement approach on-orbit.

pressed air thrusters. Each vehicle has onboard com-

Consider the case shown in Fig. 2, which corre-

puting and batteries, and communicates with the

sponds to the situation for the ground based testbeds.

other vehicles via a wireless Ethernet, making them

The unknown state of the

th vehicle is the 7(cid:2)1 vec-

i

self-contained and autonomous.

tor

= [

], where

is the position of vehicle

i

i

i

i

X

P

; E

P

American Institute of Aeronautics and Astronautics

3

Pseudolite k

P
X = =
E

i

i
i

p
p
p

ix

iy
iz
i1
i2

i3
i4

K ijk

Qk

D

ijk

f ijk

Antenna j

B

ij

Pi

World Frame

8

and 8

6=

. The

in Eq. 2 are the intra-

ijk

k

j

m

M

vehicle integers.

The inter-vehicle double di(cid:11)erences contribute

primarily to the determination of the relative po-

sitions between each vehicle. Given

pseudolites,

N

there are

(cid:0) 1 unique double di(cid:11)erences between

N

pseudolites

and

(

6=

). These di(cid:11)erences

k

k

k

k

1

2

1

2

are calculated in order to eliminate the remaining

e(cid:11)ects due to clock di(cid:11)erences

(

(cid:0)

)

c

(cid:28)

(cid:28)

1

2

v

v

Vehicle i

r(cid:1)

= j

1

2

1

j (cid:0) j

1

j

ijk

k

imk

jmk

(cid:30)

D

D

(cid:0)(j

2

j (cid:0) j

2

j) +

1

2

(3)

imk

jmk

ijk

k

D

D

(cid:21)N

8

,

with

6=

. The

in Eq. 3 refers to

ijk

k

1

2

k

k

k

k

N

1

2

1

2

the inter-vehicle integers.

The double di(cid:11)erence measurements are coupled

Fig. 2:

De(cid:12)nition of the vehicle state and GPS mea-

to the states of the two vehicles used in each pair-

surements for the indoor formation (cid:13)ying testbeds.

ing, so all of the measurements must be combined to

i

p

; p

; p

E

with coordinates (

), and

is the orien-

ix

iy

iz

i

tation of the vehicle, represented in quaternion form

(cid:15)

; (cid:15)

; (cid:15)

; (cid:15)

1

2

3

4

(

).

i

i

i

i

calculate the entire formation state. From equations

(2), (3), and the quaternion constraints (

+

+

2

2

(cid:15)

(cid:15)

1

2

i

i

2

2

(cid:15)

(cid:15)

3

4

i

i

+

= 1) , the complete set of measurements can

be related to the vehicle states:

The measured carrier phase at antenna

of ve-

j

hicle

from pseudolite

is then

i

k

(cid:1)

(

)

jk

jk

(cid:30)

1

h

X

1

1

M

1

2

3

2

3

2

3

0

(

)

0

h

X

1

c

(cid:30)

D

c(cid:28)

c(cid:28)

(cid:21)K

ijk

ijk

vi

pk

ijk

= j

j +

+

+

(1)

6

7

6

7

6

7

(cid:30)

2

h

X

2

2

M

2

(cid:1)

(

)

jk

jk

6

7

6

7

6

7

where

is the line-of-sight vector from the phase

D

ijk

center of the pseudolite antenna to the phase center

6

7

6

7

6

7

0

(

)

0

h

X

2

c

6

7

6

7

6

7

=

+

(4)

(cid:21)

6

7

6

7

6

7

(cid:1)

(

)

jk

jk

(cid:30)

3

h

X

3

3

M

3

6

7

6

7

6

7

6

7

6

7

6

7

6

7

6

7

6

7

0

(

)

0

h

X

3

c

6

7

6

7

6

7

on each receive antenna. The terms

and

c(cid:28)

c(cid:28)

vi

pi

6

7

6

7

6

r(cid:1)

(

)

k

k

7

1

2

h

X

; X

N

12

1

2

12

(cid:30)

12

k

k

1

2

represent the portion of the phase incurred by clock

4

(cid:30)

5

4

5

4

5

h

X

; X

N

r(cid:1)

(

)

k

k

1

2

k

k

1

2

23

23

2

3

23

6

7

6

7

6

7

errors between the transmitter and the receiver, and

are the dominating terms in the measurement equa-

In this equation,

is a set of nonlinear functions of

h

i

tions. These terms are eliminated by di(cid:11)erencing

X

h

i

i

i

1

2

,

is a second set of nonlinear functions of the

over measurements from multiple antenna and vehi-

given vehicle states, and

is the quaternion con-

h

c

cles. The term

is the integer ambiguity for each

K

ijk

straint function for each vehicle. For simplicity, we

antenna and pseudolite pair. The antenna location

have written the measurement equations assuming

relative to the nearby pseudolite is clearly a function

that the double di(cid:11)erences are performed between

of the vehicle position and attitude, and thus

D

ijk

sequentially numbered vehicles.

is a function of the entire vehicle state

.

X

i

Given the phase measurements for the vehicle

The intra-vehicle single di(cid:11)erences contribute pri-

formation, the optimal estimate of the formation

marily to the attitude determination for each vehi-

states

(

) can be solved in real-time using non-

X

t

i

cle. These measurements are obtained by di(cid:11)erenc-

linear estimation techniques. However, as shown

ing between the master antenna (

=

) and each

j

m

in Eq. 4 it is also necessary to obtain an estimate

of the slave antennas

of vehicle

for measurement

j

i

of the single and double di(cid:11)erence integer ambigui-

from pseudolite

k

ties so that these can be subtracted from the phase

measurements. Refs [3, 6, 19] discuss integer reso-

(cid:1)

= j

j (cid:0) j

j +

(2)

ijk

imk

ijk

ijk

(cid:30)

D

D

(cid:21)M

lution approaches based on the motion and search

techniques [20, 21]. These initial results are promis-

American Institute of Aeronautics and Astronautics

4

e
e
e
e
ing, and work continues on implementing these ap-

to increase the sky coverage.

proaches on the testbed shown in Figure 1.

Geometry:

The scale of the relative distances are

This calculation solves for the absolute positions

quite di(cid:11)erent for the indoor and LEO environments.

of the vehicles in the formation. The relative posi-

In the indoor case, the distance from the formation

tions can then be found by di(cid:11)erencing these abso-

to a pseudolite is approximately the same as the dis-

lute estimates. A CDGPS sensing system typically

tance between vehicles within the formation. This

provides a much more accurate estimate of the rel-

presents a number of issues that are discussed by

ative position between the vehicles because many of

Zimmerman [2]. One of the most important issues is

the most important error sources are predominantly

the di(cid:11)erence between spherical or planar RF signal

common-mode and their e(cid:11)ects are removed by dif-

wavefronts. In the indoor case, with a near constella-

ferencing the absolute position estimates.

tion, the RF signal wavefront must be assumed to be

3.1 On-orbit Operations

rier phases to the vehicle states (as given in Eq. 4).

spherical. This requires a nonlinear mapping of car-

To this point the discussion has focused on forma-

trial case using the far constellation) this mapping

tion (cid:13)ying techniques in an indoor environment us-

is greatly simpli(cid:12)ed because we can assume planar

For the LEO environment (or any outdoor terres-

ing pseudolites.

It is important to note how this

wavefronts.

work relates to operations in a Low Earth Orbit envi-

ronment using the NAVSTAR satellite constellation.

Environment:

The di(cid:11)erences in operating envi-

This section discusses some of the key di(cid:11)erences

ronments will have many other impacts as well. For

between the indoor and LEO environments and the

example, we would expect much less multi-path in

impact of these di(cid:11)erences on the Orion mission.

LEO. Multi-path is re(cid:13)ected GPS signal interfer-

ence which degrades the performance of the sensor.

Synchronous transmitters:

One key di(cid:11)erence

However, the large temperature (cid:13)uctuations char-

is that the indoor pseudolites are not synchronized,

acteristic of the space environment will change the

whereas the NAVSTAR satellites have extremely ac-

GPS line biases, which could degrade the sensing

curate clocks that are closely synchronized. In mov-

performance. Design of the spacecraft,

including

ing to the LEO environment, this synchronization is

placement of antennas to minimize multi-path and

very bene(cid:12)cial, as will be discussed in what follows.

thermal control, will be required to resolve these is-

Motion:

Spacecraft in a low Earth orbit environ-

sues.

ment will generally experience very di(cid:11)erent mo-

Resolving these di(cid:11)erences:

As discussed, some

tion than most terrestrial vehicles. Furthermore,

of the changes associated with moving from indoor

the NAVSTAR satellites are in motion, changing

to LEO operations can be accounted for in the de-

position relative to the LEO satellites (which dif-

sign of the spacecraft. Other changes, such as the

fers from the indoor case with (cid:12)xed position pseu-

increased Doppler shifts, the satellite motion, the far

dolites). This relative motion between the NAVS-

constellation geometry, and the synchronous signals

TAR satellites and LEO spacecraft will result in a

will result in changes to the GPS estimation algo-

large Doppler shift of the received RF signals. As

rithm and measurement equations.

discussed in Ref. [15], these large Doppler shifts sig-

ni(cid:12)cantly increase the frequency range that must be

In particular, in the LEO case, the vector

D

ijk

searched to acquire and lock onto the GPS signals.

changes, and can be considered to be only a func-

These changes impact the selection of the integer

tion of position, decoupling the attitude and position

initialization approach.

solutions. Because of the availability of a synchro-

nized signal, the position can be determined by pseu-

The LEO spacecraft may also rotate at rates and

doranging. The pseudorange equation is similar to

in directions that are not common to terrestrial vehi-

Eq. 1, with

= 0,

assumed known and small,

K

(cid:28)

ijk

pk

cles, resulting in an issue of how to keep the NAVS-

and

is the user time bias that must be calculated.

(cid:28)

vi

TAR satellites in view of the vehicle antennas. This

These pseudorange measurements are typically lin-

can be resolved using many (non-aligned) antennas

earized [13] and then solved to determine crude es-

American Institute of Aeronautics and Astronautics

5

timates (accuracy on the order of 10 m) of the ab-

need only wait and the line-of-sight directions to the

solute vehicle positions (in Earth-centered reference

GPS constellation will change su(cid:14)ciently. Fortu-

frame). The attitude of the spacecraft is calculated

nately, these initialization algorithms can be tested

using single di(cid:11)erences [15]

prior to (cid:13)ight using the hardware-in-the-loop GPS

(cid:1)

=

A

[

]

+

(5)

ijk

BE

j

ijk

(cid:30)

S

b

(cid:21)M

k

E

B

T

(cid:2)

(cid:3)

simulator available at Stanford.

TM

where

is the normalized line-of-sight vector to the

S

k

4 The ORION

GPS Receiver

k

th GPS satellite represented in the Earth-centered

frame of reference, and

is the baseline vector from

b

j

As is clear from the preceding sections, GPS sensing

the

th antenna to the Master antenna represented

j

is a key element in the autonomous navigation and

in the body frame. The direction cosine matrix A

BE

control system for the Orion formation (cid:13)ying mis-

is a function of the vehicle attitude. Note that this

sion. Thus the micro-satellites that are discussed in

single di(cid:11)erence equation is similar to Eq. 2, ex-

more detail in the next section must include (cid:13)exible,

cept in this case, given knowledge of the line-of-sight

inexpensive, but very capable GPS receivers. Sev-

vectors, the attitude calculation decouples from the

eral GPS receivers have already been developed for

vehicle position. The relative positioning between

space applications and these were compared using

the spacecraft could then be improved using dou-

a variety of criteria: physical size, power required,

ble di(cid:11)erences, once again resulting in centimeter-

number of channels, space experience, degree of fa-

level accuracies. The procedure for calculating the

miliarity at Stanford, cost, access to the source soft-

double di(cid:11)erence equations is the same as the one

ware, and performance. Based on this analysis, it

used to develop Eq. 3 from Eq. 2. Note that in this

was clear that none of the available receivers was

case the time information available from the NAVS-

a good match in terms of the most important fea-

TAR transmitted data can be used to time tag the

tures: power, code access, and cost. Thus the deci-

measurements on di(cid:11)erent spacecraft, resulting in

sion was made to develop an in-house GPS receiver

synchronization accuracies on the order of 10’s-100’s

that would be expandable, modular, and have an

nanoseconds.

open software architecture. These goals were met

by modifying the Mitel GPS ORION

chipset

.

TM

2

Initialization:

The approach to determining po-

sition and attitude on start-up will also change in

Hardware:

The ORION design was modi(cid:12)ed to in-

the LEO environment. The (cid:12)rst issue is the in-

corporate dual front ends on a single circuit board,

creased Doppler shifts. Without any modi(cid:12)cation,

with a capability to receive an o(cid:11)-board clock signal

the time required to acquire a particular satellite’s

to drive the receiver circuitry. These modi(cid:12)cations

signal could be longer than the total time the satel-

allow for an arbitrary number of dual front-end re-

lite is in view, potentially resulting in the situation

ceiver cards to be chained together using a common

that a position solution is never acquired. This prob-

clock. In this con(cid:12)guration any number of antennas

lem can be remedied by including an orbit propaga-

can be used to form a GPS based attitude system.

tor (also required for the dynamic state estimator).

A four antenna receiver has been constructed us-

This propagator would estimate the satellite’s posi-

ing this approach. The ORION lower board has 3

tion and velocity, as well as estimate the NAVSTAR

separate serial connections on-board. Each receiver

satellite’s position and velocity (based on the last

card has one dedicated link, while one serial port is

recorded almanac data). With this information, the

shared by both cards. Although each receiver card

frequency search space would be greatly decreased,

has its own processor (ARM-60), the dedicated se-

thus reducing the time to acquire the GPS satel-

rial links can be used by the receiver boards to share

lites [15]. Another di(cid:11)erence in the initialization is

information. This allows for a distributed processing

the resolution of integer cycles in the attitude and

environment wherein the CPU’s share the work load

relative position equations. Unlike the (cid:12)xed pseudo-

associated with the higher-level relative navigation

lites, the NAVSTAR satellites are in motion, chang-

and control algorithms.

ing the relative line of sight to the user formation.

Software:

A key bene(cid:12)t of using the Mitel GPS

Therefore the initialization would not require addi-

tional motion from the user vehicles, because they

2

Name should not to be confused with the mission name.

American Institute of Aeronautics and Astronautics

6

Phase A

Phase B

Phase C

hardware is that this also provides access to the

GPSBuilder Software. This software is entirely writ-

ten in C code. Complete access to the code has al-

lowed us to extensively modify the code and carrier

tracking loops, the signal acquisition algorithms, the

cycle-slip detection routines, the input/output rou-

tines, and the frequency search region during startup.

Testing:

Several tests have been performed to de-

termine the contribution to the di(cid:11)erential carrier

phase (DCP) measurement noise due to the clock

synchronization method. The DCP was measured

using a zero antenna baseline, and the receive an-

Fig. 3

: Basic phases of the Orion (cid:13)ight mission.

tenna stationary and moving. A splitter was used

receiver, this design should meet the needs of the

to route the antenna signal to all four RF inputs on

the receiver and integrated carrier phase was tracked

Orion (cid:13)ight mission.

for approximately two minutes at a 1 Hz data rate.

Measurements were made of the integrated carrier

5 Orion Mission Overview

phase from an array of indoor pseudolite transmit-

ters in a high multipath environment. The results

for the stationary case showed that the error in the

DCP measurement between tracking channels on the

same RF input is very small (STD 0.2 mm), and is

within the resolution of the integrated carrier phase

measurement. When DCP measurements are made

using two RF front ends on the same card, the error

grows to a STD of 1.5 mm. And for a DCP measure-

ment between two RF front ends on separate cards,

the STD of the measurement error is only 2.2 mm.

These tests indicate that it is possible to make DCP

measurements using multiple boards synchronized at

the 10 MHz clock level and obtain errors that are

comparable to single board receivers (such as the

Trimble TANS Vector used on the formation (cid:13)ying

testbed in Figure 1).

The ob jective of the Orion mission is to demon-

strate several key sensing and autonomous control

technologies that are necessary to develop a virtual

spacecraft bus. This will be accomplished using a

distributed array of simple, but highly coordinated

micro-satellites designed and built in-house. As il-

lustrated in Figure 3, the current plan is to use a con-

stellation of six satellites to demonstrate the relative

ranging techniques. The satellites will be launched

and deployed in one or two stacked sets (

). This

A

con(cid:12)guration will be used to perform an initial refer-

ence calibration of the GPS receivers. The next step

will be to split the stacks and perform coarse sta-

tion keeping of the micro-satellites within each trio

(

) (possible scenario: 1 km separation with toler-

B

ances of approximately 100 m). The vehicles will be

controlled within an error box and 3-axis stabilized

The results for the moving antenna case are com-

using feedback from the onboard GPS receiver.

parable, but slightly higher (3-4 mm) than the sta-

tionary case. The error is slightly higher due to the

presence of two cycle slips during the collection pe-

riod. These slips occurred when the antenna was

moved quite rapidly from side to side, but the cy-

cle slip detection algorithm was able to detect and

correct for the integer wavelength jump in the DCP.

When we have determined that the six satellites

are functioning properly, the two groups will be com-

bined into a single coarse formation. The next phase

(

) will be to perform precise station keeping ma-

C

neuvers for periods of approximately 1/2 an orbit

(possible scenario: 100 m along track vehicle sepa-

ration controlled to approximately (cid:6) 5 m tolerance

This modi(cid:12)ed ORION receiver is currently in

along-track and radial). The real-time relative sepa-

use on the 3D blimp testbed, and a similar device

ration and attitude measurements will be validated

has been tested in a LEO environment on the GPS

using onboard cross checks between the six vehicles.

simulator. The results from these tests are very

A simple digital camera will be used to verify the

promising, and the indication is that, with some ba-

pointing accuracy within the formation. More so-

sic changes to improve the radiation tolerance of the

phisticated real-time validation techniques, such as

American Institute of Aeronautics and Astronautics

7

Fig. 4

: SQUIRT Satellite

Fig. 5

: Actual Sapphire satellite hardware.

low while still achieving a 6-12 month operational

laser ranging, will be included as permitted by the

life.

power and mass budgets. The primary means of val-

idating the real-time measurements will be to store

To meet the mission goals, the micro-satellites

and then downlink the raw carrier phase and pseudo-

will need both attitude control and station keeping

range data. Measurements will be taken on-orbit

ability. For autonomous operation, a high perfor-

while selected ground stations have the same GPS

mance inter-satellite communications system would

constellation in view. Down-linked data will then

also be required. The ma jor e(cid:11)ort to date has fo-

be post-processed to validate the real-time measure-

cused on determining the general requirements for:

ments using techniques already demonstrated on the

JPL TOPEX mission.

1. Size, weight and power capability of the bus

During all phases of the mission, the commands

structure,

from the ground will specify maneuve