Journal of Physics: Conference Series

PAPER • OPEN ACCESS You may also like

Accuracy improvement of inertial navigation by

- Spacecraft autonomous navigation using
unscented particle filter-based
celestial/Doppler information fusion Xiaolin
special methods of measurements using Ning and Jiancheng Fang

microelectromechanical systems - The method of increasing the accuracy of

coordinates of the contaminated area and
water surfaces during environmental
monitoring
To cite this article: A P Voyskovskiy et al 2019 J. Phys.: Conf. Ser. 1353 012031 M E Kulikova, A I Nagornaya and M V
Diuldin

- The application of displaced trajectories in

tasks of aircraft navigation in time
V V Markelov
View the article online for updates and enhancements.

This content was downloaded from IP address 103.153.131.253 on 15/11/2022 at 05:25

HIRM-2019 IOP Publishing
Journal of Physics: Conference Series 1353 (2019) 012031 doi:10.1088/1742-6596/1353/1/012031

Accuracy improvement of inertial navigation by special

methods of measurements using microelectromechanical
systems

A P Voyskovskiy, D M Kruzhkov, V A Yakimenko and N V Kartushina

Moscow Aviation Institute, 4, Volokolamskoye Avenue, Moscow, 125993, Russia

E-mail: [email protected]

Abstract. The article is devoted to the description of algorithmic mechanism devised for
increasing accuracy of inertial navigation based on the use of low-priced (up to 1000 usd)
microelectromechanical measuring systems installed on small and mini-aircraft. The proposed
algorithms include a series of error model parameters updating operations, which are used to
negate inaccuracy occurred upon conducting measurements and their further utilization. It is
shown that not only the algorithms proposed by the authors are capable to massively increase
accuracy of navigation in comparison with the unprocessed initial navigational information
obtained from the measuring devices, but also allow using of low-priced
microelectromechanical systems for brief autonomous navigation by integrating aircraft
reference motion equations.

1. Introduction
Modern civil and military aircraft are usually equipped with a huge variety of navigation gauges,
including inertial measuring devices – optoelectronic hyroscopes and accelerometers, which are
considered to be the basis of strapdown inertial navigation system (SINS) [1, 2]. These devices allow
high accuracy of measurements whilst maintaining autonomous navigation. However, due to high cost
and unavailability to common consumers such equipment is not utilized neither on small and amateur
aircraft, nor on modern UAVs. Cheap and small aircraft are usually equipped with inertial measuring
devices, produced with high-sensitivity microelectromechanical components [3]. They are featuring
low price, variety, relatively simple setup and possibility to integrate with the aircraft softwawre for
navigational needs. At the same time there are a number of issues preconditioned by the peculiarities
of its utilization, which are: substanial errors, measurement noise, measurement error parameter drift
[4], all of which leads to the accumulation of critical navigation errors upon attempting to integrate
motion equation and calculating reference trajectory onboard solely by SINS in autonomous mode. All
these factors lead to a severe decrease in the possibility of practical utilization of low-priced
microelectromechanical systems (MEMS) onboard.
Therefore, cheap MEMS have prospects of them being used on manned aircraft only under the
condition of them enjoying higher accuracy and reliability. Without cost increase and necessity to
integrate with other onboard measuring devices [5, 6] this goal can only be reached by researching
specialized algorithms and software, for example [4, 7], which are capable of conducting intelligent
adaptive processing of initial navigation data and updating error models of measuring of a specific
device with the considered accumulated statistics. It is obvious that in order to maintain constant
increase of navigation measurments accuracy throughout target function – flight – it is necessary for
the software

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
HIRM-2019 IOP
Journal of Physics: Conference 1353 (2019) Publishing

bearing to work simultaneously with the algorithms for navigation measurements analysis above stated
functions. In order to create such algorithm it is essential to elaborate the nature of processes, which
lead to the deterioration of accuracy of navigation measurements, conducted by inertial MEMS, then
draw up adequate models of errors with vatious nature. It is now necessary to look into the authored
thesis through the example of software designed to increase accuracy of navigation solution with
accelerometers manufactured by "Analog Devices" company.

2. Methods
Mathematical model of errors occuring in various microelectromechanical measuring devices is
formed through the analysis of their operating principles and related documentation by modifying an
alternate version presented in [8]:
a  a0 t   a1 t, at   K1 at   K2 at   a2 u,t  (1)
 a3 t 
being t – moment of measuring, at – true value of acceleration (real), K1 – nonlinearity factor of
accelerometer transfer-function, K2 – quotient of accelerometer scale range error consideration, a0 –
zero shift, a – random process describing behavior of the device's noise, a – a set of harmonics
1 2
characterizing device's noises caused by vibrations and other external fluctuations but not linked to the
useful signal a , u – parameter-function to define input control, which amplifies
t
microelectromechanical device measurement errors during flight, a3 – a measurement error "outbreak"
outside of the area characterized by the density of measuring error distribution. It is necessary to point
out, that such model is not complete, as it does not include a number of components, for example,
related to non-orthogonality of measuring device sensitive axles etc.
The analysis of the formula (1) allows pointing out several traits of accelerometer measuring errors,
which can be represented as a random process or a random value. If they are not eliminated when
resolving navigation task, ultimate error becomes rather substantial, and the attempt to integrate
motion equation will shortly (up to 5 mins) lead to an exponential divergence of the solution. To avoid
such divergence it is necessary to utilize adaptive mechanisms of initial navigation data processing as
to eliminate fixed and random measurement errors before the data is integrated and used in the
navigation system itself. It is necessary to add, that engine start and initiation of the flight cause such
massive disruption of measuments conducted by cheap microelectromechanical hyroscopes and
accelerometers, that using such measurements for navigation calculations to integrate vessel trajectory
becomes literally senseless. Therefore, software is to be designed and tested statically right before the
flight, as well as dynamically after engine start and ascent. The authors are going to look into main
calculations, performed by the authors with initial navigation data received upon aircraft's pre-launch
activation, in order to evaluate error parameters and models which occur when the engine is already
started and therefore accelerometer is receiving substantial vibration interference.

definition of zero shift a0 (1). Zero bias estimate is conducted through the processing of
selection of measurements and evaluation of mathematical expectation. Its value is then deducted from
the initial navigation data which allows to negate this fixed error [5]. After processing a map of shifts
for the aircraft's static and pre-launch activation is drawn up, which as well allows to evaluate zero
shift's divergence confidence interval.

definition of zero drift a1 (1) and forming the fixed error trend of measurements. The trend
parameters are selected for three variants of approximation: linear, square and cubic, selected on the
basis of minimal ultimate dispersion of the initial navigation data drift on specific timeline relating to
the chosen model of the trend. The length of the functional timeline is defined experimentally on the
basis of the trend's behavior and zero drift evaluation – statically for a specific model of the device. In
order to accumulate corresponding statistics it is necessary to test-run the two modes – aircraft static
after engine start but before the flight and the flight itself. The drift parameters are evaluated on the
basis of Kalman estimator [9] by processing external navigation data, which allows updating
2
HIRM-2019 IOP
Journal of Physics: Conference 1353 (2019) Publishing
parameters of

3
HIRM-2019 IOP
Journal of Physics: Conference 1353 (2019) Publishing

a1 model (one of three kinds) during flight for their further use when external data is unavailable. The
form of approximating function is selected on the basis of minimal error of polynomial extrapolation
relating to zero drift values calculated subsequently onboard.

eliminating gravitational acceleration projections on the axles of measuring devices [8]. This
operation is conducted by drawing up a system of equations reflecting connection between calculation
in X, Y, Z channel and the angle of high-sensitivity device base (it is considered that the device axles
are orthogonal), occurred due to the noncollinearity between gravitational acceleration vector and
vertical axle of the accelerometer, that is why the stated error is not included directly in (1), but still
exists.

outbreak evaluation and their elimination upon processing of measurements. "Outbreaks" are
calculation errors, which are irrelevant to the realities and exceed certified error values of a specific
measuring device defined by the manufacturer. Upon exceeding measurement estimates by K·2·σ (σ
being sum of noise error and fixed error standard deviations, K – confidence factor, by default equals
1). Isolated measurements are therefore excluded from the further calculations and processing.

engine vibration noises filtering a2 (1) is implemented by processing initial navigation data with
the use of first-order aperiodic filters, whose cutoff frequency is defined by the type of powerplant
installed on the aircraft and spectral density of the noise, caused by the fluctuations of moving parts.
Due to the spectral density being both a function of time and of the aircraft parameters during flight
(which is predefined by the existence of transition modes in engine operation), a more complex noise
filtration a2 is conducted takes into account reference motion model type and control model type. Just
as a1, noise parameter update (the type of approximating processes spectral density) during flight is
implemented with the specific traits of the aircraft taken into account by accumulating statistics and
processing external calculations from other sensors in the onboard algorithm.
All the values of the device errors obtained through the above stated procedure are deducted from
the calculations of the accelerometers, which are then transferred to the input of navigation system.

3. Elaborated method efficiency estimation

In order to evaluate the efficiency of the drawn-up algorithms and software a number of experiments
were held, centering around consecutive resolution of navigation task in the channel of the aircraft
vertical motion (ascend), during which initial navigation data is recorded and processed on every
iteration with the use of the above stated algorithms. According to the results, continuous update of the
model (1) quotients and parameters allowed substantially increasing the accuracy of the integration to
implement aircraft autonomous navigation. The example of algorithms in use is shown in Figure 1.
During experiments the authors found out, that altitude evaluation error in vertical channel in static
with engines online increased two-fold and reached about 1 cm after 5 mins of the systems operation
with the use of cubic polynomial, which approximates accelerometer zero drift, 20 cm without
updating device zero drift, and 1 m without any use of the algorithms stipulated in section 2.
The experiment results during flight are given in Figure 2. The figure shows the evolution of the
aircraft altitude error during brief autonomous flight featuring multiple low-amplitude ascend and
descend. Dash line stands for integrating measurements without use of algorithms. The other two
curves show the trajectory with the use of the algorithms stipulated in section 2, whose
implementation is characterized either by the use of linear or of polynomial interpolation of zero drift
in high-sensitivity devices. The application of these algorithms without interpolating device zero drift
never lead to adequate results and is close to behavior of the dash-line relation.

4
HIRM-2019 IOP
Journal of Physics: Conference 1353 (2019) Publishing

Figure 1. Navigation errors. Without use of algorythms – dash line; using algorithms without
polynomial to describe zero drift – dotted line; using all of the algorithms – dot-dash line

Figure 2. Navigation errors. Without use of algorithms – dash line; using all of the algorithms as well
as cubic polynomial to describe zero drift – dotted line; using all of the algorithms as well as linear
polynomial to describe zero drift – dot-dash line

4. Conclusion
It is shown that the algorithms devised by the authors are capable to substantially (two-fold and
higher) increase accuracy of navigational sighting whilst having low-priced inertial measurement
devices installed on small aircraft. Therefore, the proposed algorithms and software allow
implementing autonomous navigation over 5 mins without any support on behalf of other navigation
systems, including satellites, solely basing on the use of cheap inertial devices with high-sensitivity
MEMS’ components. Moreover, these algorithms can be subject to various upgrades, including their
possible adaptation to particularities of specific hardware installed onboard of a real aircraft. A built-in
mechanism of evaluating inherent accuracy and the level of the result probability belief can be viewed
as a rational supplement to the above stated algorithms. Another aspect of their future improvement is
adaptation to the use in an aircraft's maneuver flight modes.

References
[1] Gao W, Zhang Y and Wang J 2014 Sensors 14(1) 1511–1527
[2] Tawk Y, Tome P, Botteron C, Stebler Y and Farine P-A 2014 Sensors 14(2) 3768–3796

5
HIRM-2019 IOP
Journal of Physics: Conference 1353 (2019) Publishing

[3] Sun W, Wang D and Xu L 2013 Mean.: J. Int. Meas Confed 46(8) 2585-2596
[4] Gao W, Zhang Y and Wang J 2015 Sensors 15(2) 3154–3171
[5] Veremeenko K K and Savel’ev V M 2013 J. Comp. Sys. Sci. Int. 52(1) 106–116
[6] Golovan A A and Nikitin I V 2015 Moscow Univ. Mech. Bull. 70(2) 46–49
[7] Kim M, Yu S-B and Lee K-S 2014 Int. J. Prec. Eng. Manuf. 15(3) 567–575
[8] Fang B, Chou W and Ding L 2014 Int. J. Adv. Robot Sys. 11(1) 14
[9] Simon D 2006 Optimal State Estimation: Kalman, H∞, and Nonlinear Approaches (Cleveland:
John Wiley and Sons)