Aerospace and Electronic Systems Magazine July 2018 - 20

A Survey on Quadrotors
and integrates them progressively into the expertise with several known variables. It also performs the "weighted fusion" of
influence values in the construction of the controller. In [100],
the authors have designed and implemented a Fuzzy Logic Controller. Other researchers have presented an intelligent fuzzy
logic controller to control the position and the orientation of the
quadrotor [101].

Model Based Predictive Control (MBPC)
MPC controllers are designed based on the dynamic model of the
system that must be controlled, using mathematical optimization
techniques to obtain optimal inputs to be applied to the system.
Essential aspects are accuracy of the dynamic model and the calculation load of the optimization to be executed. According to
[78], the MBPC is an advanced control technique, essentially a
process of repeated optimizations and constraints at each time
step. The quadrotor is a nonlinear system, so it requires a nonlinear MBPC controller. The MBPC optimizes the trajectory at
each time; a path optimization unique reference provides state
values error. The closed loop control can be obtained by continuous re-optimization of the path. There is no guarantee of computing time, and the MBPC controller depends on the accuracy of
the model; therefore, if the model changes due to, for example, a
change in mass, the resulting control input may not be suitable.
The presented MBPC uses the quad rotor differential dynamics to
achieve the overall trajectory generation and control. This system
has been validated using a complete quadrotor model developed
using experimental data and theoretical analyses to accomplish
three flight missions. Such a system can be used to improve the
behavior of autonomous UAVs. The results show the benefits of
using MBPC to achieve autonomous flight, including the ability
to re-optimize the trajectory at each time to reflect environmental
changes [81].

Adaptive Control Algorithms
Adaptive controllers are aimed to adapt uncertain or time varying
parameters with changes in the system. In [102], a continuous time
varying adaptive controller is presented. This controller performs
well with known uncertainties in mass, moments of inertia, and
aerodynamic damping coefficients. In [103], an adaptive controller
using a feedback linearization is implemented for quad rotors with
dynamic changes in the center of gravity. It is noted that when the
center of gravity changes, the adaptive controller is able to stabilize the system; PD and regular feedback linearization techniques
are not sufficient to stabilize the system. In [104], an adaptive control technique based on rectilinear distance is used with a tradeoff
between control performance and robustness. The modified (linearized) model was able to compensate for constant and moderate
wind gusts.

Robust Control Algorithms
Robust controller algorithms guarantee controller performance
within acceptable disturbance ranges or unmodeled system parameters. The main limitation of this type of controllers is poor tracking ability. In [105], a robust controller is implemented and vali20

dated experimentally for an attitude controller subsystem based on
linear control and robust compensation. In [106], a robust tracking
algorithm is presented. This strategy shows asymptotic stability in
the presence of parametric uncertainties and unknown nonlinear
disturbances.

IDENTIFICATION OF A QUAD ROTOR MODEL AND STATES
ESTIMATION
System identification, or the art of obtaining mathematical models of physical systems from the measured input-output data, has
been developing since 1965. However, many open problems persist. References [107], [108] present such problems in cases of
quadrotor nonlinearity and closed loop identification. It is worth
noting that important attention has been given to the identification
of helicopters, rather than multirotor systems, due to the relative
complexity in dynamics, as shown in [109]. Some strategies developed for quadrotor model identification include:
Parameters Estimate Based on First Principle Model, as presented in [29], [110], [111]. Several aerodynamic parameters can
be obtained by regular identification methods [112], [113], [114].
Masses, rotational inertias and displacements, and motor constants
as presented in [115], [116]. Reference [112] presented the z inertia
and the rotor parameters obtained by applying the Levenberg-Marquardt optimization method and a quadratic optimization method
the.
Steel-Grey Models: Local Linear Models Identification, as
shown in [117]; for the time domain identification as shown in
[118]; the frequency domain identification as presented in [119];
and the identification based on the Unscented Kalman Filter UKF
as presented in [120].
Slate-Grey Models: RBF-ARX Model as shown in [121].
Black Models: Data-Based Model as presented in [122]-
[126].
Reference [21] presents several newly-published works on the
modeling and parameter identification of a quadrotor, while [34]
and [127] present a survey for the categorization of quadrotors
based on identification approaches. Reference [128] proposes a
higher order sliding mode observer for states and parameters estimation and identification.

COLLISION AVOIDANCE SYSTEMS (CAS)
In [129], the authors present a collision avoidance systems (CAS)
that uses a radar sensor at every time frame. Then, the trajectory
of the quadrotor is realized according to the region with the lowest
return. It is worth noting that the quadrotor needs to keep tracking
its heading and position every time it performs a maneuver in order
to determine when it returns to the original track. If a continuous
collision-free path does not exist, the quadrotor will be lost. Other
research studies present a CAS that is oriented towards probabilistic collision detection methods. The type of sensor used is automatic dependent surveillance-broadcast, and the goal is to detect
other aircraft (obstacles); the maneuver approach is chosen and
pre-defined according to different probabilities of collision. This
strategy requires huge processing capacity in order to simulate all

IEEE A&E SYSTEMS MAGAZINE

JULY 2018

Table of Contents for the Digital Edition of Aerospace and Electronic Systems Magazine July 2018