Aerospace and Electronic Systems Magazine September 2017 - 4

Feature Article:

DOI. No. 10.1109/MAES.2017.160194

Design for Graceful Degradation and Recovery from
GNSS Interruptions
Trevor Layh, Demoz Gebre-Egziabher, University of Minnesota-Twin Cities,
Minneapolis, MN, USA

INTRODUCTION
In guidance, navigation, and control (GN&C) of small unmanned
aerial vehicles (UAVs), estimates of the vehicle's kinematic states
are generated by an integrated navigation system. In current applications, the system of choice is an inertial navigation system (INS)
aided by measurements from a global navigation satellite system
(GNSS) receiver. One of the shortcomings of these integrated
GNSS/INSs is the problem of temporary or prolonged GNSS outages. These outages can occur because of temporary signal loss due
to obstructions, a prolonged outage due to interference or jamming,
or deliberate action by the GN&C system to isolate a failed receiver
or reject an anomalous signal in space. In these instances, the position, velocity, and attitude solutions generated by processing the inertial measurement unit (IMU) outputs alone in INSs quickly drift.
To mitigate this drift, alternate aiding signals such as cameras [1]-
[5], radars [6], light detection and ranging (LIDAR) [7], or other
signals of opportunity [8] have been used. When the only kinematic
state of interest is attitude (e.g., UAV stabilization), IMUs aided by
magnetometers and airspeed sensors have been used to mechanize
attitude heading reference systems (AHRSs). A system architecture that uses an AHRS and airspeed measurements to mechanize
a dead-reckoning (DR) navigator aided by the relative range measurement between cooperating vehicles is discussed in [9], [10].
Designing integrated navigation system algorithms that operate effectively in both GNSS-available and GNSS-denied environments is rather complex. At least three key challenges must be dealt
with in designing such algorithms. First, it is necessary to design
fault detection and isolation algorithms that simultaneously have
low false alarm and missed detection rates-a conflicting set of
requirements. Second, because fault detection algorithms cannot
detect the onset of a fault instantaneously, there is a short window
when a failed GNSS receiver or anomalous signal in space is used
to aid the IMU. Once the fault has been detected, removing or unwinding its effect from the inertial navigation solution is not trivial.
Third, the transition from GNSS-available to GNSS-denied (and
Authors' current address: T. Layh, D. Gebre-Egziabher, University of Minnesota, 110 Union Street SE, Minneapolis, MN
55455, USA, E-mail: (gebre@aem.umn.edu).
Manuscript received September 7, 2016, revised and ready for
publication November 30, 2016.
Review handled by M. Braasch.
0885/8985/17/$26.00 © 2017 IEEE
4

vice versa) operational modes must occur quickly and smoothly.
This is particularly true of the attitude solution, because it is indispensable for UAV control. Jumps or discontinuities in the attitude
solution can lead to situations in which the flight control system
(FCS) generates commands that destabilize the UAV. Graceful degradation is a term sometimes used to describe systems that have
features allowing them to deal with some or all of these challenges.

OBJECTIVE
The purpose of this article is to present a decentralized filtering approach to design an integrated navigation system for small UAVs
that seamlessly switches between GNSS-available and GNSS-denied modes of operation. When GNSS services are denied, it gracefully degrades to a less optimal operational mode while maintaining
sufficient accuracy to allow continued guidance and control of a
small UAV. When GNSS services are restored, it smoothly transitions back to a high-accuracy operational mode. The decentralized filter presented in this article is based on the idea of fusing
the output from a set of parallel filters. Each parallel filter can be
a stand-alone system that provides an estimate of all or a subset of
the vehicle states needed for UAV GN&C. As shown and discussed,
this approach can easily be implemented on most existing UAV
FCSs. This is because it does not require additional sensors beyond
those already found on most FCSs on the market today. The filter
addresses the second and third challenges noted in the preceding
paragraph, namely, the ability to unwind the effect of a failed sensor
and provide smooth transitions to and from GNSS-denied operations. Finally, the performance of the proposed filtering approach is
validated using data from flight tests of a small UAV.

PRIOR WORK
The body of literature discussing decentralized filtering in general
is vast and goes back several decades [11]-[17]. In what follows,
we limit our review to literature in which decentralized filtering is
applied to enhancing the robustness and flexibility of integrated
navigation systems. A lucid but high-level treatment of the decentralized filtering for integrated navigation systems can be found
in [13]. A thorough list of references on the topic can be found in
[13] and [14]. While the presentation in [13] discusses the merits
and drawbacks of several decentralized filtering architectures, neither results from case studies nor rigorous analysis assessing the

IEEE A&E SYSTEMS MAGAZINE

SEPTEMBER 2017

Table of Contents for the Digital Edition of Aerospace and Electronic Systems Magazine September 2017