Graphics Reference
In-Depth Information
2
20
0
10
-2
0
40
45
50
55
60
40
45
50
55
60
time [s]
time [s]
2
1.4
1
1.2
0
1
-1
40
45
50
55
60
40
45
50
55
60
time [s]
time [s]
Fig. 4.5 Acceleration and velocity in x ( blue ), y ( green )and z ( red ) when throwing the MAV in
the air. At the throw, the MAV experiences an acceleration of 16.5 m/s 2
( top left ). The angular
velocities when throwing the MAV rise to 190 /
s( top right ). And the velocity rises up to 2.3 m/s
( bottom left ). The scene depth is estimated correctly at all times ( bottom right ) and that the MAV
can maintain it drift free after convergence. This allows good initialization of a subsequent, more
powerful VSLAM algorithm
4.3 Obstacle Detection and Avoidance
The ability to sense obstacles and avoid collisions autonomously is a critical safety
component for MAVs flying in cluttered environment and at low altitude (Fig. 4.6 ).
Driven by payload and processing constraints and the requirement to accurately detect
every obstacle within collision range, new algorithms have to be developed that reflect
the limited capabilities of such small platforms. In the following section, we present
our obstacle avoidance subsystem that uses stereovision for range perception, and a
polar-perspective, inverse-range world representation (disparity space) for collision
checking. The perception system is used by a closed-loop RRT motion planner, to
navigate around detected obstacles, incorporating vehicle dynamics.
4.3.1 Related Work
Significant progress has been made recently on deliberative obstacle avoidance using
active optical range sensors, such as single-axis scanning lidar and RGB-D structured
light [ 5 , 56 ]. The lack of a second axis for scanning lidar is a significant limitation
and structured light sensors are ineffective in sunlight. Very compact, electronically
beam-steered radar with large maximum range is under development for this appli-
cation, but again has a single-axis scan [ 52 ]. Optical flow can cover a wide field
of view and has been used for reactive obstacle avoidance [ 14 , 25 ]; however, this
 
Search WWH ::




Custom Search