www.sea-technology.com May 2016 / st 35
A
utonomous navigation is an im-
portant task for USVs. Robot navi-
gation methods involve solving the
path-planning or obstacle avoidance
problem and optimizing a path with
respect to certain criteria. Obstacle
avoidance comprises global and lo-
cal obstacle avoidance. Many scholars
have carried out extensive research on
path planning with various algorithms.
In local path planning, the APF
method is a typical path-planning
method for USVs. Although it is very
suitable for low-level, real-time control
of the robot due to its explicit physical
parameters and simple mathematical
description, there are some defects in
the original algorithm. The accuracy of
obstacle avoidance is affected by the
measuring errors of lidar.
This article focuses on solving
problems of the defects of the APF al-
gorithm and sensor errors. We divide
USV navigation into two parts: ac-
curate obstacle avoidance and fuzzy
obstacle avoidance. First, the USV will avoid ob-
stacles with the modifed APF algorithm in the ac-
curate obstacle avoidance part. Second, the USV
will be controlled by the fuzzy obstacle avoidance
algorithm when the measuring errors of lidar are
too big.
Data Filter
We use SICK's LMS511 lidar in the obstacle
avoidance module. The resolution of the LMS511
aspect angle is 0.5°, and the scanning angular
scope is from -5° to 185°. However, there exist er-
rors in scanning data due to the effects of environ-
ment and the USV's motion. Thus, it is necessary to
flter original data to eliminate all unreliable data
Autonomous USV
Obstacle Avoidance
Integrated Algorithm to Solve APF, Sensor Defects
By Peng Wu • Dr. Shaorong Xie • Hengli Liu
(Top) Simulation result in the general area. (Bottom)
Simulation result in the local minimum area.