You are on page 1of 6

2017 14th International Conference on Electrical Engineering, Computing Science and Automatic Control (CCE), Mexico City, Mexico.

September 20-22, 2017

The Hexapodopter: a holonomic flying hexapod


Daniel Soto-Guerrero and José Gabriel Ramı́rez Torres
Cinvestav Tamaulipas
Cd. Victoria, Tamaulipas, México
dsoto, grtorres at tamps.cinvestav.mx

Abstract—This document introduces an air-land robotic plat-


form called Hexapodopter, capable of ground locomotion and
vertical take-off and landing (VTOL). For walking, each of the
limbs feature three degrees of freedom (DoF); for flying, the
thrust is provided by a set of motors and propellers mounted on
every limb. Due to its mechanical design, the Hexapodopter can
reconfigure the thrust vectors of propellers into a true holonomic
configuration during flight; in other words, the vehicle is capable
of controlling individually its 6 degrees of freedom on the air,
without reorienting any of the thrust motors nor the body. In
this document, we present a formal analysis of the holonomic
properties of the vehicle, not found in related works, as well
as a control architecture for the Hexapodopter, exploiting these
properties, shown through numerical test results.

Fig. 1: A render view of the Hexapodopter.


I. INTRODUCTION
Hexacopters are a multicopter variant featuring six sets
An holonomic aerial vehicle, capable of moving in any
of motors and propellers, radially disposed around the main
direction without changing its current orientation, could effec-
body, to generate the lift force required for flying. Also, an
tively counteract external disturbances coming in any direc-
hexapod robot is a very stable ground vehicle, supported
tion. Thus, in such a vehicle, the translation motion and the
by six electrically-powered limbs. The Hexapodopter, the
orientation pose of the robot can be decoupled through the
vehicle we propose in this document, is a hybrid of the two
control law, leading to a more robust and simpler controller
vehicles just mentioned. The cohesion of the two, complement
architecture.
themselves to make the vehicle an holonomic flying vehicle
with walking abilities (see Fig. 1). The Hexapodopter is an effective holonomic aerial robot,
capable of flying in any direction without tilting and because
Mechanically, a multicopter consists of several brushless
it has six limbs with 3 DoF, it is capable of legged locomotion
motors which, under a control law, provide the thrust force to
[15]. On this document we will focus on formally proving and
achieve an stable flight. The most common arrangements use
testing its holohomic performance when airborne; to do so we
4, 6, or 8 parallel motors perpendicularly fixed to the body
used the well-known condition of Force Closure [3], used to
frame; these arrangements guarantee an even distribution of
prove grasping stability in robotic hands.
the weight of the vehicle between all thrust motors, maximiz-
ing flight times. This standard arrangement has been heavily
studied and effective control laws have been proposed for it II. R ELATED WORK
[1], [2].
A major drawback to the standard arrangement is that the A multicopter can be considered as a rigid body with
vehicle can only move sideways by first leaning the main body six degrees of freedom whose state can be described by
towards the desired direction of movement. This is achieved by six independent variables: 3 for its position and 3 for its
increasing the angular speed of one of the motors and decreas- orientation. Therefore, when airborne, the thrust forces and
ing the angular speed of its diametrically opposite, creating a torques exerted by the motors on the multicopter’s body frame
torque τ around a given axis and causing the orientation of should suffice to control the 6 degrees of freedom of the
the multicopter to change and gain horizontal speed. Since it vehicle.
is not possible for the vehicle to move horizontally without Several different ways to arrange the motors of an un-
first changing its orientation, the parallel motors arrangement manned multicopter have been proposed with the purpose to
forms a constraint on the state speeds of the vehicle, called increase their maneuverability. They all fall in two distinctive
nonholonomic constraint. This constraint makes impossible categories: fixed and dynamic arrangements. What differenti-
for multicopters to counteract external horizontal disturbances ates fixed from dynamic arrangements is whether or not they
(e.g. the disturbance caused by the wind) without leaning first. actively re-orientate the thrust forces during flight.
978-1-5386-3406-6/17/$31.00
c 2017 IEEE
2017 14th International Conference on Electrical Engineering, Computing Science and Automatic Control (CCE), Mexico City, Mexico. September 20-22, 2017

db Main body diameter.


lb Distance along the ZB axis, from OB to OL .
ω1
dp Diameter of the propeller.
ω2
lm Height of the brushless motor.
lc Length of the coxa.
lf Length of the femur.
θ2
θ1
lt Length of the tibia.
β2 Body frame lp Distance along XL axis from OL to OP .
ω3 β1
B XB ω4
YB
ZB
TABLE I: Dimensional parameters of the hexapodopter.
β4
θ3 θ4
β3
w Kinematic chain to the propeller
N
W D-H Parameter
E n Description
D
d θ a α
1 0 θs 0 0 The swing servo
π
Fig. 2: Thrust vectoring quadcopter. 2 d2 0 lc 2
Coxa
3 d3 θl − π2 lp − π2 Fixation of the motor
4 hm 0 0 0 Propeller’s location (P )
Kinematic chain to the support joint
A. Fixed arrangements D-H Parameter
n Description
On fixed arrangements, all the motors, and therefore the d θ a α
thrust forces they generate, are fixed with respect to the center 1 0 θs 0 0 The swing DoF.
π
2 d2 0 lc The length of the coxa
of mass of the multicopter. The fixed arrangements are the 2
3 d4 θl − π2 lf 0 The lift DoF
most common in multicopters featuring four [6] or six [2] 4 d5 −θk lt 0 The knee DoF, tip of leg (SP )
parallel motors. All these vehicles are nonholonomic systems,
requiring relatively complex controller architectures to achieve
TABLE II: Denavit-Hartenberg parameters for every limb.
stable hovering and flight.
Shimizu et al. proposed an hexacopter with three non con-
secutive motors tilted inwards and the rest tilted outwards [12]. hexacopter, makes the hexapodopter a holonomic multicopter
Similarly, Rajappa et al. proposed an optimization technique that falls into the dynamic arrangements category.
to find the best orientation for each motor of an hexacopter
before every flight, so the control effort for a given trajectory
III. D ESCRIPTION OF THE H EXAPODOPTER
can be minimized [9]. Although they did not proved formally
the holonomic characteristic of these arrangements, they did Figure 1 shows a render view of the vehicle we are propos-
managed to operate independently on the 6 DoF of the vehicle. ing, a hybrid vehicle between hexapods and hexacopters. A
brushless motor is fixed to provide lift force on every 3 DoF
limb and all limbs are radially and evenly distributed around
B. Dynamic arrangements the main body frame (see Figs. 3). The Swing articulation
Arrangements falling on this category can actively (during is capable of protraction and retraction. The Lift articulation
flight) reorient the thrust forces with respect to the center of provides depression and elevation (Dep/Lev), while the Knee
mass of the multicopter. The simplest dynamic arrangement articulation allows for flexion and extension (Flx/Ext). The
is a tricopter that can tilt one of the three motors through brushless motor is fixed on the Femur of the vehicle and
a steering mechanism; the axis of rotation of the steering therefore, all thrust forces can be re-orientated with respect
mechanisms is coaxial with the radius (the θi DoF in Fig. to the center of mass B of the robot.
2) [13]; the same approach can be migrated to a quadcopter, The location of the i-th leg and its corresponding mounting
except that the four motors are capable of tilting [8]. Taking point Mi , relative to B, is described with a rigid body
this approach one step further, the thrust motor can be mounted transformation matrix M,i B A; given by the rotation matrix
M,i 3
on a 2 degrees-of-freedom steering mechanism, so it can move B R ∈ SO and the translation vector M,i tB as:
in two perpendicular directions (θi and βi DoFs on Fig. 2), this M,i M,i

configuration is also called thrust vectoring. All the previously M,i
A = BR tB
B
mentioned dynamic arrangements are capable of increasing 0 1
maximum acceleration by profiting from gyroscopic effects Starting at every mounting point Mi , the i-th limb can be
and differential thrust at the cost of being over-actuated and modeled as a kinematic chain using the Denavit-Hartenberg
yet, they remain as nonholonomic vehicles because the thrust (DH) parameters [4]. Given the 8 physical dimensions of the
vectoring can only provide accelerations in only 4 directions, vehicle, listed on Table I, the corresponding DH parameters
that certainly can be reoriented dynamically. for the i-th limb are shown on Table II.
On comparison to the related works, we proved formally There are two points of interest on every limb: for flying,
the holonomic characteristic of the hexapodopter and, based the location of the i-th thrust motor Pi and for legged loco-
on the numerical results, we designed the control architecture motion, the location of the i-th support point SPi , both with
for the vehicle. Furthermore, the cohesion of the hexapod and respect to B. Given the DH parameters, for every link we
2017 14th International Conference on Electrical Engineering, Computing Science and Automatic Control (CCE), Mexico City, Mexico. September 20-22, 2017

(−1)i kτ ω 2 , where kf and kτ are proportionality constants [14]


ω3
and (−1)i defines the sign to take into account the rotation
τ2 ω2
sens of the propeller. From here, every motor exerts a force
τ3 and torque, with respect to the body frame, given by:
∅dp

lf
fi = kf ωi2 n̂i
τ i = kf P,i tB × n̂i + (−1)i kτ n̂i ωi2

ω4 XB lc τ1

θs 60

Z Y where n̂i is the unitary vector coaxial with the ZP,i axis of
τ4 OB B BOS,1 OL,1
ω1
OK,1
∅db
lp the motor. Both, the force fi and torque τ i , can be grouped in
a wrench wi :
   
τ6 fi n̂i
kf ωi2
 
wi = = P,i i kτ
ω5 τ5 τi B tB × n̂i + (−1) kf n̂i
ω6  
Tf
wi = k ω 2 = Ti fi
Tτ i f i
(a) The robot base of the Hexapodopter. with fi = kf ωi2 being the magnitude of the thrust force and
Swing Servo
Ti being the transfer colon vector for motor i. Therefore, the
θs Brushless motor total force and torque applied to the six thrust motors, with
ZS,i ωi
ZM,i
Lift Servo
YS,i YL,i
respect to B, can be obtained as follows:
M,i θl ZP,i
OB A Knee Servo
YB B
XS,i
 
Femur YP,i YK,i
Body frame
XB OM,i Y OL,i F
ZB
XM,i M,i
OP,i XP,i = TΩ
Coxa ZL,i XL,i
XK,i lm
τ B
lp ZK,i OK,i
lc
θk where:
lf  
Tibia lt
Tf  
T= = T1 T2 · · · T6

ZSP,i
T
Ω = kf ω12 ω22 · · · ω62
OSP,i

XSP,i
YSP,i
Support point
Using the transfer matrix T, the dynamics of the
(b) The 3 DoF limb of the Hexapodopter. Hexapodopter can be rewritten as follows:
Fig. 3: Mechanical description of the hexapodopter. mr̈ = R (q) Tf Ω − mg − mq̇ × ṙ (3)
Iq̈ = Tτ Ω − q̇ × Iq̇ (4)
P,i
can compute the homogeneous transformations M,i A(θs , θl )
SP,i
and M,i A(θs , θl , θk ). Said that, the location of the i-th V. H OLONOMIC A NALYSIS
thrust motor, relative to B, is given by P,i M,i P,i
B A = B A M,i A
If the Hexapodopter is a 6 DoF vehicle, the rank of T
and the location of i-th support point can be computed by should be equal to 6, but this is not sufficient. To describe
SP,i M,i SP,i and validate that the Hexapodopter is a holonomic UAV we
B A = B A M,i A.
used the force-closure concept.
IV. M ATHEMATICAL MODEL FOR FLYING
In this section, we present the mathematical model for the A. Force-Closure in Grasp theory
Hexapodopter during flight, which is based on the Newton- Force-closure has been used to validate that a mechanical
Euler equations and determined by the physical parameters hand can grasp an object while maintaining translational and
shown in the past Section. The relative position and pose of B rotational equilibrium [3]. We used this idea to validate that the
with respect to W is given by the position vector r = [x, y, z]T six thrust forces grasping the body of the Hexapodopter while
and the three Euler angles q = [φ, θ, ψ]. flying, can counteract the forces and torques caused by any
The dynamics of a rigid body can be decomposed into external disturbance. In grasping theory, a force acting on an
two components: translational and rotational, as shown in object on a contact point Ci is modeled by a four-component
Equations 1 and 2, respectively. vector fi = [fn , fo , ft , fm ]Ti , where the component fn is the
normal component, fo and ft are the two perpendicular tangent
mr̈ = R(q)F − mg − q̇ × mṙ (1)
components and fm represents the spin moment about the
Iq̈ = τ − q̇ × Iq̇ (2) contact normal. The exerted wrench by a force fi on a rigid
with m the total mass of the rigid body, I its inertial matrix, g body is modeled with a grasp matrix Gi ∈ R6×4 and the
the gravity acceleration, R(q) the rotation matrix, F the total resultant wrench can be computed as wi = Gi fi , where Gi is
force and τ the total torque applied over the rigid body. defined by:
To relate the angular speed ωi of the motor with the thrust
 
n̂ ô t̂ 0
force it generates, we do f = kf ω 2 and the torque as τi = Gi =
r × n̂ r × ô r × t̂ t̂ i
2017 14th International Conference on Electrical Engineering, Computing Science and Automatic Control (CCE), Mexico City, Mexico. September 20-22, 2017

Configuration np rank(Gn ) d>0


where n̂i is the normal vector to the surface of the rigid body
on the contact point Ci while ôi and t̂i are perpendicular Quadcopter [6] 4 4 7
Thrust vectoring quadcopter [5] 4 4 7
tangent vectors to the surface, forming the orthonormal frame Tilting propeller quadcopter [11] 4 4 7
Ci on the contact point. The frames Ci and B are related by Hexacopter [1] 6 4 7
the position vector. C,i tB . Tilting propeller hexacopter [10] 6 6 4
If J forces are exerted on a rigid body we can test the 6 DoF hexacopter [12] 6 6 4
Fixed vectoring hexacopter [9] 6 6 4
equilibrium condition of the rigid body by computing and
Hexapodopter 6 6 4
expecting the resultant wrench p to be equal to zero. Then, the
equilibrium equation is p = Gn f , where Gn is the grasping
TABLE III: Results of the LP1 test. Only those configurations
matrix given by:
with rank(Gn ) = 6 and d > 0 are form-closure.

Parameter Value Parameter Value


Gn = [G1 , G2 , . . . , GJ ] ∈ R6×4J
kf 4.0e-6 Kg·m lc 0.06 m
f = [f1T , f2T , . . . , fJT ]T ∈ R4J×1 kt 3.3e-8 Kg· m2 lf 0.16 m
θS 0 rads lp 0.16 m
If a grasp configuration satisfies p = 0, then it is a steadily db 0.12 m Fmax 6.85 N
grasp; on which case, there are two possible situations, the
TABLE IV: Configuration parameters for the hexapodopter.
grasp is either force-closure or form-closure. The first, denotes
a successful grasp but implies that there is the possibility of
geometrical movement on one of the DoF of the grasped ob-
ject, retained by frictional forces. The second case constitutes the body frame, and the augmented grasping matrix Gn will
a successful grasp, but the grasped object has no geometrical be:  
chance to change any DoF.   Tf mgB
Gn = T wB =
In order to obtain a form-closure grasp (or a force-closure Tτ 0
grasp), the resultant wrench p must span R6 under the The adapted LP problem can be stated as follows:
restriction that all finger forces fi are applied against the
object. Mathematically, the form-closure grasp has a first-order LP1: maximize d
equivalent statement: there exists a vector λ n such that subject to: Gnλ n = 0
rank(Gn ) = 6, Gnλn = 0 and λn > 0. This is commonly
λn − 1d ≥ 0

stated as an optimization problem:
d≥0
LP: maximize d
λn = 1
subject to: Gnλ n = 0
λn − 1d ≥ 0
Iλ The condition λn = 1 indicates that the last component of
vector λn , representing the weight, is a constant force.
d≥0 We tried the LP1 test with different configurations found
where the solution d gives an idea about the strength of the on the literature, the results are shown on Table III. As
grasp. expected, the standard quadcopter and hexacopter are not form
closure. The Hexapodopter and other related works proved to
be form closure. The LP1 test proofs formally the holonomic
B. Form closure test and Holonomic UAV
characteristic or the hexapodopter, i.e. np +1 forces make out a
By applying grasping theory to the Hexapodopter by mod- form-closure grasp around the UAV, therefore we can conclude
eling the thrust motors as friction-less contact points with that any external disturbance wrench can be compensated and,
proportional spin moments: inversely, any controlled wrench can be applied to the UAV.
   
fn Ω
= kτ
fm i kf Ω i C. Optimal configuration
we find that the transfer matrix T and the grasping matrix Gn After proving formally the holonomic characteristic of
are equivalent. the hexapodopter, we aimed to find the optimal configu-
An interesting fact about UAV control theory is that, in ration for flying that provides the best grasp, defined by
general, the weight mgB of the vehicle is considered a [θL,1 , θL,2 , . . . , θL,6 ]T . We chose the optimal configuration
perturbation to be compensated by the controller with the n based on the d term from LP1, which is defined in grasp
thrust forces. In our approach, the weight is considered as an theory as the smallest finger force required to maintain the
additional constant force, that can be exploited as a grasping object’s equilibrium before any wrench is applied. It represents
force. the maximal disturbance, applied in the same direction of that
Because T and Gn are equivalent, we reformulated the LP specific finger, that the grasp can compensate before losing
problem to include the wrench wB = [mgB , 01×3 ]T exerted contact with the object. Thus, the parameter d measures the
by the weight of the UAV as a grasping force. For a UAV with grasp quality. Since the finger forces are positively unbound,
np thrust motors, there are n = np + 1 wrenches applied over the bigger the value d, the better the grasp is.
2017 14th International Conference on Electrical Engineering, Computing Science and Automatic Control (CCE), Mexico City, Mexico. September 20-22, 2017

Parameter Value
12 Mass (m) 1.6 Kg
Inertia (IB ) diag(2.5e − 2, 2.5e − 2, 5.1e − 2)Kg · m2
0.25
10 θL 40o

0.2 TABLE V: Simulation parameters for the hexapodopter.

Acceleration [m/s2 ]
8
d
Force [N]

Γ(d) 0.15
6
ẍmax Controllers Modelled
ÿmax dynamics r, ṙ
4 0.1
fd uF
rd , ṙd Position Translational
Control F Vehicle
2 0.05 ωB,d uτ allocation dynamics
qd , q̇d Attitude Rotational
q, ωB
0 0
0 10 20 30 40 50 60 70 80
Lift angle θL [Deg]
Fig. 5: Controller architecture. The position and attitude con-
Fig. 4: Maximum acceleration in the x and y axes, versus the trollers are decoupled, acting independently.
form closure metric d and Γ(d).
Figure 4, the shaded area spans the peak of performance
Since the LP1 test considers the weight of the vehicle, the function Γ(d) to the peak of maximal accelerations. We
term d represents the smallest thrust force required to coun- consider that any lift angle in the shaded area represents
teract its weight. Under the restriction that ωi ∈ [0, ωmax ], all a good compromise between grasp configuration of thrust
motors can exert a maximum thrust force Fmax ; if d . Fmax , motors (providing a better stability during flight to compensate
then the controller would have a small scope for action and external disturbances) and maneuverability of the UAV.
the vehicle would be consuming too much energy just to
compensate its own weight. Said that, we propose a quadratic VI. P OSITION CONTROL SYSTEM
function Γ(d) to define the compromise between grasp quality After analyzing and validating the holonomic characteristic
and maneuverability, given by: of the Hexapodopter, we aimed to design a closed loop
controller for it. The chosen architecture for the controller
Γ(d) = d(Fmax − d) is depicted on Fig. 5. This controller was numerically tested
to perform a challenging flight plan: take-off, describe a
The function Γ(d) serves as a performance metric for a lemniscate and land. The lemniscate, consisting of two turns
given configuration of the Hexapodopter, defining a compro- on opposite directions and two straight line segments, requires
mise between the grasp quality d and the scope for action from the vehicle to move in every direction. The parametric
Fmax − d of the UAV. equations for the lemniscate at a flight altitude h are:
Through simulation, we evaluated Γ(d) for the finite range
0 ≤ θL ≤ 90◦ , using the simulation parameters from Table IV
a sin( t )
   
and considering ωmax = 12500RP M . The results are shown φd (t) = 0
in Fig. 4, the function Γ(d) increases lightly from 0◦ to around rd (t) =  b sin( 2t )
 qd (t) =  θd (t) = 0
4t
40◦ , where it reaches its maximum value, and then decreases c sin(  ) + h ψd (t) = 0
rapidly until reaching 0 at around 68◦ . When Γ(d) is negative, The error between the desired and current position, orien-
the motors are turning at ωmax , but they cannot compensate tation and velocity of the vehicle at time instant k, defined by
the weight of the Hexapodopter. ek = rd − r(tk ) is fed to the control architecture, both for
Since classical multicopters cannot move sideways without position and attitude control, which are two PD controllers.
tilting, we also computed the maximal acceleration in x and y The translational and rotational dynamic equations 3 and 4
T
directions (on the XYW plane), that the Hexapodopter can are used to compute the required wrench [uf , uτ ] while
achieve for a given lift angle θL without tilting, in order the transfer matrix T (the Control Allocation block) is used
to analyze the maneuverability of our design [7]. These two to compute the desired angular speeds for the thrust motors
subproblems can be stated as follows: (see Fig6b). Additional simulation parameters are listed on
Table IV. The initial conditions used in the simulation were
r(0) = [0, 0, 0]T and all Euler angles equal to zero.
ACC1: max ẍ subject to: ÿ = z̈ = q̇ = 0 The executed flight plan is shown in Fig 6a. The graphic
F
ACC2: max ÿ subject to: ẍ = z̈ = q̇ = 0 shows how the vehicle takes off at ground level and per-
F
forms the desired trajectory given by rd (t), then it lands at
The results for ACC1 and ACC2 are also shown in Fig. 4. r = [0.5, 0, 0]T . Figure 6c displays the changes in altitude
As expected, the maximal accelerations increase from 0 at and the resultant angular velocities for all the motors can be
θL = 0◦ to their corresponding peaks at around 55◦ . observed on Fig. 6b, all velocities being positive and bounded
These two metrics, the performance function Γ(d) and the ∈ [0, ωmax ] during the whole trajectory. Figure 6d shows the
maximal accelerations (ẍ, ÿ), can be used to determine the evolution of the Euler angles; Roll and Pitch and Yaw angles
best configuration for the lift angle of the Hexapodopter. In stay close to zero, i.e. the UAV remains leveled at all times.
2017 14th International Conference on Electrical Engineering, Computing Science and Automatic Control (CCE), Mexico City, Mexico. September 20-22, 2017

rd (t) performance of the Hexapodopter without interfering with its


r(t)
ability to perform legged locomotion.
1.5

Z axis [m]
R EFERENCES
1
[1] V.S. Akkinapalli, G.P. Falconi, and F. Holzapfel. Attitude control of
a multicopter using l1 augmented quaternion based backstepping. In
0.5 Aerospace Electronics and Remote Sensing Technology (ICARES), 2014
IEEE International Conference on, pages 170–178, Nov 2014.
0 [2] A. Alaimo, V. Artale, C. Milazzo, A. Ricciardello, and L. Trefiletti.
1 Mathematical modeling and control of a hexacopter. In Unmanned
0.5 1
0 0.5 Aircraft Systems (ICUAS), 2013 International Conference on, pages
0
−0.5 −0.5 1043–1050, May 2013.
−1 −1 [3] ZUO Bingran and QIAN Wenhan. A force-closure test for soft multi-
Y axis [m] X axis [m]
fingered grasps. SCIENCE CHINA Technological Sciences, 41(1):62,
(a) The desired and actual trajectories of the 1998.
vehicle, after executing the flight plan. [4] Peter Corke. A simple and systematic approach to assigning denavit-
hartenberg parameters. IEEE Transactions on Robotics, 23(3):590–594,
June 2007.
Angular velocity [Rad/s]

960 [5] Minh-Duc Hua, Tarek Hamel, Pascal Morin, and Claude Samson.
ω1 ω2 ω3 ω4 ω5 ω6
940 Control of vtol vehicles with thrust-tilting augmentation. Automatica,
920
52(C):1–7, February 2015.
[6] A. Kaba, A. Ermeydan, and E. Kiyak. Model derivation, attitude control
900
and kalman filter estimation of a quadcopter. In 2017 4th International
0 50 100 150 200 250 300 350 400 450
Conference on Electrical and Electronic Engineering (ICEEE), pages
Time [s]
210–214, April 2017.
(b) The angular speeds of the 6 motors. [7] H. Mehmood, T. Nakamura, and E. N. Johnson. A maneuverability
analysis of a novel hexarotor uav concept. In 2016 International
Conference on Unmanned Aircraft Systems (ICUAS), pages 437–446,
Roll angle φ [Rad]

1.5 June 2016.


1 [8] A. Nemati and M. Kumar. Modeling and control of a single axis tilting
0.5 rd (t) quadcopter. In 2014 American Control Conference, pages 3077–3082,
0 r(t) June 2014.
0 50 100 150 200 250 300 350 400 450 500 [9] Sujit Rajappa, Markus Ryll, Heinrich H. Bülthoff, and Antonio Franchi.
Time [s] Modeling, control and design optimization for a fully-actuated hexarotor
aerial vehicle with tilted propellers. IEEE International Conference on
(c) The altitude response during the whole flight.
Robotics and Automation (ICRA), May 2015.
[10] M. Ryll, D. Bicego, and A. Franchi. Modeling and control of fast-hex:
·10−13 ·10−12 ·10−14
a fully–actuated by synchronized–tilting hexarotor. IEEE/RSJ Int. Conf.
Pitch angle θ [Rad]

4 2
Yaw angle φ [Rad]
Roll angle φ [Rad]

1 on Intelligent Robots and Systems, 2016.


3
1 [11] M. Ryll, H. H. Bülthoff, and P. R. Giordano. A novel overactuated
2 0 quadrotor unmanned aerial vehicle: Modeling, control, and experimen-
1 tal validation. IEEE Transactions on Control Systems Technology,
−1 0
23(2):540–556, March 2015.
0 [12] T. Shimizu, S. Suzuki, T. Kawamura, H. Ueno, and H. Murakami.
0 200 400 0 200 400 0 200 400
Time [s] Time [s] Time [s] Proposal of 6dof multi-copter and verification of its controllability. In
Society of Instrument and Control Engineers of Japan (SICE), 2015 54th
(d) The Euler angles for the vehicle. Annual Conference of the, pages 810–815, July 2015.
[13] Ziwei Song, Kunpeng Li, Zhihao Cai, Yingxun Wang, and Ningjun Liu.
Fig. 6: The simulated flight of the vehicle. Modeling and maneuvering control for tricopter based on the back-
stepping method. In 2016 IEEE Chinese Guidance, Navigation and
Control Conference (CGNCC), pages 889–894, Aug 2016.
[14] T. Tomic. Evaluation of acceleration-based disturbance observation for
VII. C ONCLUSIONS multicopter control. In Control Conference (ECC), 2014 European,
pages 2937–2944, June 2014.
In this paper, we presented a hybrid novel robotic platform [15] H. Yu, H. Gao, L. Ding, M. Li, Z. Deng, and G. Liu. Gait generation
called Hexapodopter, which is the union of a VTOL multirotor with smooth transition using cpg-based locomotion control for hexapod
walking robot. IEEE Transactions on Industrial Electronics, 63(9):5488–
machine and a legged robot. We focused on formally describ- 5500, Sept 2016.
ing the holonomic flying performance of the vehicle, which to
the knowledge of the authors, was not be found elsewhere.
Based on the analysis of the holonomic performance, we
proposed the metric Γ(d) to evaluate a suitable configuration
for the robot and a closed-loop control architecture to make it
execute a challenging flight plan.
We have proven mathematically that the ability to change
the orientation of every thrust force makes the vehicle holo-
nomic, in the sense that it can perform flying maneuvers,
such as a lemniscate, while maintaining roll and pitch angles
≈ 0, i.e. without tilting. Because locomotion of hexapods have
been proven successful before [15], we can conclude that
the synergy of the two robotic platforms increase the flight

You might also like