The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008 
needed to scan the whole area displayed in figure 6. Both 
measurements are computed in the inertial frame. 
Figure 7. Helicopter position along inertial frame x axis. 
Figure 8. Helicopter yaw angle along the reference trajectory. 
Figure 9 displays in green color the GPS position, computed 
along the x axis, returned by the measuring sensor block on the 
first 100 seconds of simulation. These noisy measurement is 
compared with the values of the reference trajectory (blue lines) 
and with those representing the current state as output by the 
helicopter dynamic model (noiseless). Then in figure 10 the 
output of the EKF for the roll angle is shown overimposed on 
similar data. Therefore this figure summarizes the whole block 
scheme developed for the autonomous control of our model 
helicopter. The blue line represents the reference trajectory, the 
red curve is the current UAV state (for the roll angle) which is 
undirectly “seen” by the control system through the noisy 
measurements output by the Mti-G unit, shown in green. The 
latter are entered in the Kalman filter in order to obtain a better 
tracking (light blue line) of the reference trajectory. 
Figure 11. Roll angle: reference, state, measurements and EKF 
output.- 
Finally figures 12 and 13 shows an example of the trajectory 
tracked in Simulink by the control system in terms of the 
inertial position along the x axis and of the yaw angle for the 
first 400 seconds of simulation. The time shift for both curves 
with respect the theoretical ones is due to a typical side effect of 
control systems based on filtered measurements. Here both 
curves (in red) are time delayed by 1 second. Achieved results 
show that during a scan line (figure 6), our model helicopter 
would shift correctly along the y axis but the yaw angle tends to 
produce a translational components even along the x axis. 
However in this case major displacements occur during motion 
between stop points: here the control loop correctly moves the 
helicopter on the reference trajectory so that the image capture 
position is very nearly to the reference one. 
In order to evaluate the performance of implemented control 
system we computed the RMS error for all the 6 variables 
defining the helicopter position and attitude, that is: 
? J N 
e vXMS = ' X £y ’ (14) 
where s v (t) = v # (t) - (t ) 0</< 400 s 
Figure 12. Actual and reference trajectory along the x-axis 
component of helicopter motion. 
Figure 9. GPS position along x axis with added noise.