Вы находитесь на странице: 1из 3

L3 A PPLICATIONS : G ROUP 6 F INAL R EPORT

Andrei Hirjanu, Viktor Hiscock, Tian Ci Hong, Alexander Hunter, Joseph Ibrahim,
Jonathan Jackson, Benhur Johnson

L3Group6.weebly.com
his report summarises the work of
Group 6 in the L3 Applications Exercise at Imperial College London. A
jump-gliding robot no larger than 40 cm in
size and weighing no more than 50 g, was to
be designed, built and tested. Performance
of the robot is measured by the distance attained. Our final model was a dart-like construction that achieved a range of 1.6 m.

launch speed could be achieved, shown in figure 2.


Lastly, the script attempts to account for mechanical
and voltage losses by applying conservative efficiency
factors based experience in past application exercises.

3 Glide Model

Since the glide distance is of the order of the launch


altitude, it follows that the robot will be in a transient
1 Introduction
state for the majority of its trajectory. Hence the only
The three pillars of science and engineering are theo- reasonable assumptions that can be made regarding
retical analysis, mathematical modelling and experi- the dynamics of the robot are:
mentation; implicit in this is the subject of optimisa- 1. Longitudinal and lateral motions are uncoupled
(Curie Principle)
tion. This represents an iterative procedure whereby
the results of experimentation are fed into the math- 2. Elastic deformations are negligible (rigid body
motion analysis)
ematical model for refinement; the approach utilised
3.
Ground effects are negligible
in this exercise.
The process was initiated by idealising the physical Assumption 1 is valid provided that the states of
problem into something more readily amenable to the system are not close to those that would induce
analyse. Hence the trajectory was split into two a bifurcation in the state space. Assumption 2 is
distinct phases: (1) Drive Phase (2) Glide Phase.
necessary to make at this stage since aeroelastic
behaviour can only be evaluated once initial geometry
has been defined. Assumption 3 is valid since ground
2 Drive Model
effects at the low flight speeds considered here are
The drive model guided the selection of an optimal indeed negligible.
motor and estimated the launch speed - a critical
The free body diagram in figure 3 illustrates the
input for glide modelling. The script converts torque force system acting on the robot during the glide
to a linear force, where kinematics equations were under the aforementioned assumptions. By summing
applied to obtain the instantaneous velocity. Follow- forces in the xEarth and yEarth axes and summing moing the linear relations shown in figure 1, the output ments about the zEarth axis, Newtons 2nd Law and
torque was then updated and the process was iter- Eulers Formula may be used to obtain the governated over the ramp length while constantly checking ing equations of motion; these are also summarised
wheel traction.
in equations (4a) to (4f) on page 3. From vectorAn optimal torque could be produced by vary- geometric arguments alone, it is possible to obtain
ing the voltage across the motor at different stages analytic expressions for the remaining flight variables,
of acceleration. This meant a significantly higher some of which are displayed below for reference.

Page 1 of 3

y Earth
x Earth

The front wheel drive provides longitudinal sta(1)


bility on the ramp and also to improve the torque
!
transfer from the motor to the wheels. Given that

y Earth + (XAC,W B XCG )cos()


W B = tan1
the weight is a limiting factor, a simple tail-dragger

x Earth (XAC,W B XCG )cos()


was used at the rear of the robot as it reduces the
(2) weight without adding excessive friction.
!

Bevel gears were selected to accommodate the


y

(X

X
)
cos()
AC,W
B
CG
Earth
T = tan1
positioning
of the motor and also to improve the

x Earth (XAC,W B XCG )cos()


mechanical
stability
of the connection. The motor,
(3)
electronics and tail components were positioned such
The 2D lift, drag and moment coefficients were that the center of gravity would be situated just
each modelled using a third-order polynomial fit aft of the wheel axis to ensure good traction and
to xfoil data to account for the dependency on to avoid flipping in mid-air. Final adjustments inangle of attack. These coefficients were then used cluded installing metal bearings that reduced the
to calculate lift, drag and moment forces via their friction between the wheel axis and the chassis and
usual definitions.
adding push-fit joints to the chassis to remove the
human error associated with manually gluing small
components.
4 Combined Mathematical Model,
1

= tan

Simulation & Optimisation


By combining the two models, a system of ODEs
representing a 20-parameter optimisation problem
was created. The optimisation parameters consisted
of all dimensions in figure 3 and the wing airfoil section. This was solved within Matlabs optimisation
toolbox. Upon running this combined algorithm the
optimum trajectory in figure 4 was obtained.
The optimum configuration predicted by the model
was a wingless design. Also shown in figure 4 is a simulation for a winged design with all other parameters
held fixed. It is clear that for this launch speed the
difference in the maximum range for both winged and
wingless designs is almost negligible. This seemed
reasonable since at the low launch speed of 1.7 m/s
predicted by the Drive Model in section 2, the slightly
greater lift generated by the presence of a wing would
be offset by the accompanied increase in drag. Moreover, the model has neglected 3D effects such as
induced drag and tip losses; the losses due to the
presence of a wing would therefore be significantly
higher than those predicted by the model. Therefore
a wingless design was opted for.

6 Performance Discussion &


Evaluation

Following manufacturing, the actual speed of the


robot in a preliminary test was measured at 3 m/s
instead of the predicted 1.7 m/s. This suggests that
the loss-efficiency factors during the drive-phase were
greatly overestimated and therefore underestimated
the launch velocity for input to the Glide Model in
section 3. Figure 4 illustrates the optimum design for
a launch velocity of 3 m/s (which is indeed a winged
design) and superimposes the analogous simulation
for the actual (wingless) design tested on the ramp.
Clearly the wing provides a significant benefit at
these launch speeds. Unfortunately, Group 6 had
not anticipated such a result and, as a result, missed
the opportunity to make the design flexible enough
such that the addition of a wing and relocation of the
centre of gravity could be readily achieved following
manufacturing. Put another way, the iterative loop
described in section 1 could only be traversed a finite
number of times and hence convergence was not
achieved. Despite this, the simulation of the actual
design in figure 5 is consistent with the experimental
result obtained on test day. Hence the models were
5 Vehicle Design & Mechanical
indeed fit for purpose, but a lack of design flexibility
Implementation
and pessimistic estimates for motor efficiency were
Having extracted the optimum dimensions from the the deciding factors in final performance.
mathematical model it was now necessary to make
the following modifications due to practical considerations. Taking into consideration the optimisation
results, a narrow, lightweight chassis and a 2-wheel
front drive configuration was chosen.

Page 2 of 3

Figure 1: Torque-velocity relation at constant motor


output. Torque is updated at each instantaneous velocity.

Figure 2: Torque-velocity relation with varying output


voltage to achieve greatest launch velocity.

Translational and rotational equations of motion:


X

FxE = M x
E

1
x
E =
LW B sin(W B ) + DW B cos(W B )
M

+ LT sin(T ) + DT cos(T )

(4a)

(4b)

FyE = M y
E

1
x
E =
LW B cos(W B ) DW B sin(W B )
M

+ LT cos(T ) DT sin(T ) g

(4c)

(4d)

MCG = I

1
=
(XAC,W B XCG )LW B cos()
I
(YAC,W B YCG )LW B sin()

(4e)

+ (XAC,W B XCG )DW B sin()


+ (YAC,W B YCG )DW B cos()
+ (XAC,T XCG )LT cos(T )
(YAC,T YCG )LT sin(T )

Figure 3: Free-body diagram detailing all forces acting


on robot during glide phase.

+ (XAC,T XCG )DT sin(T )



+ (YAC,T YCG )LT cos(T ) + MAC,W B

0.8

0.8

0.6

0.6

y(m)

y (m)

WINGLESS
WING

0.4

0.4
0.2
0

(4f)

0.2

WINGLESS
WING
0

0.5

1
x (m)

1.5

Figure 4: Optimum winged (red) and wingless (blue)


designs predicted by model at 1.6 m/s.

2
x (m)

Figure 5: Optimum winged (red) and wingless (blue)


designs predicted by model at 3.0 m/s.

Page 3 of 3

Вам также может понравиться