3D Path Planning and Stereo-based Obstacle Avoidance
for Rotorcraft UAVs
Stefan Hrabar
CSIRO ICT Centre
1 Technology Court, Pullenvale 4069
Queensland, Australia
Stefan.Hrabar@csiro.au
Abstract— We present a synthesis of techniques for rotorcraft
UAV navigation through unknown environments which may
contain obstacles. D* Lite and Probabilistic Roadmaps are
combined for path planning, together with stereo vision for
obstacle detection and dynamic path updating. A 3D occupancy
map is used to represent the environment, and is updated on-
line using stereo data. The target application is autonomous
helicopter-based structure inspections, which require the UAV
to fly safely close to the structures it is inspecting. Results
are presented from simulation and with real flight hardware
mounted onboard a cable array robot, demonstrating successful
navigation through unknown environments containing obstacles.
Index Terms— UAV, autonomous helicopter, power line in-
spection, stereo vision, obstacle detection, path planning
I. INTRODUCTION
The use of Unmanned Aerial Vehicles (UAVs) is becoming
increasingly widespread, especially in military applications.
We are interested in furthering the use of UAVs in civil ap-
plications, and in particular for airborne structure inspections
(e.g., inspecting power lines, pipelines, cooling towers and
bridges). Traditionally, aerial inspections of power lines are
carried out with manned helicopters, a costly and dangerous
exercise. Williams et al [1] give a number of reasons (besides
safety) why rotorcraft UAVs are well-suited to power line
inspections. They also highlight the ability to sense obstacles
in the environment as one of the biggest challenges in using
UAVs for this task. The UAV is required to fly at close
quarters to the structure it is inspecting, and is therefore
at risk of a collision. The problem is particularly hard for
inspections carried out beyond line-of-sight of the UAV
operator, as is the case when inspecting long power lines.
The power line inspection task also requires the UAV to fly
to a goal or a number of subgoals, for example to visit a set
of transmission towers which have been roughly surveyed.
Since the precise locations of the towers and other obstacles
in the environment are not known a priori, the UAV would
need to detect these as it flew to the goal, and potentially
modify the planned path.
We detect obstacles by using stereo vision to build a 3D
occupancy map. Path planning is done using Probabilistic
Roadmaps (PRMs) [2], with D* Lite [3] to search for the
shortest collision-free path. The roadmap is updated based on
occupancy information stored in the occupancy map. Tradi-
tionally, the same 3D grid is used for occupancy mapping and
planning. The proposed technique however, permits the use
of a high resolution occupancy map with a lower resolution
planning graph, reducing the planning state space and cost.
Since we utilize existing techniques for sensing, map
building and path planning, our contribution is not so much
to these individual fields, but rather one of combining these
techniques in a novel way to facilitate 3D navigation of
a rotorcraft UAV in unknown environments. This paper
describes how we integrate these techniques, and presents
insights learned in the process. Results are presented from
experiments in 3D simulation and on the CSIRO Air Vehicle
Simulator (AVS) cable array robot.
In Section II we review the related work in this area, in
Section III we detail the stereo-based occupancy mapping
technique and in Section IV we outline the path planning
technique. Experimental results are presented in Section V
and conclusions are drawn in Section VI.
Fig. 1. CSIRO / UAV Vision mini-helicopter flying close to a power line
to demonstrate its potential inspection capability.
2008 IEEE/RSJ International Conference on Intelligent Robots and Systems
Acropolis Convention Center
Nice, France, Sept, 22-26, 2008
978-1-4244-2058-2/08/$25.00 ©2008 IEEE. 807