Sea Technology

JUL 2013

The industry's recognized authority for design, engineering and application of equipment and services in the global ocean community

Issue link: http://sea-technology.epubxp.com/i/142532

Contents of this Issue

Navigation

Page 39 of 71

Obstacle Avoidance There are two range sonars, forward and backward looking, mounted on the vehicle. Therefore, by 180-degree horizontal rotation, a 2D profle image of the water tank can be obtained. If the tank wall information is assumed to be known, then additional objects (or obstacles) in the tank can be detected through this sonar scanning. Range Sonar Modeling. Usually, it is diffcult to get satisfactory sonar measurement due to frequent failure or multipath effect, especially in the case of a small sonar beam attack angle. For P-SURO, the sonar measurement is modeled as a Gaussian distribution with the standard deviation of: Where λ1 and λ2 are sonar-dependent parameters derived through various water tank tests. The sonar attack angle is denoted by: According to this standard deviation, we determine the obstacle detection threshold: Acquisition of a 2D profle image and obstacle detection. with an online calibration process, a heading value can be obtained with less than 1 degree of drift. Windows Embedded CE 6.0 is applied in near real time to operate the embedded system, which further consists of three single board computer-based modules for vision, navigation and control. The modules connect to each other through TCP/IP ports. The real-time layer of the software frame is a thread-based multitasking structure. According to their accessing mechanism, sensor devices can be classifed into two types of active sensors that automatically output measurement and a passive one that only outputs after being requested. The passive sensor measurements as well as analog sensors use a timer routine, while individual interface threads are designed for each of the active sensors. Vision-Based Localization Visual localization methods usually can be classifed into two types. One type is based on the natural feature points of the environment, and the other type uses artifcial landmarks. The KIRO water tank is surrounded by fat concrete walls, and it is diffcult to extract specifc feature points, so artifcial landmarks were applied for vision-based localization. For P-SURO, underwater vision provides the vehicle its location for use in autonomous navigation, instead of being used to develop novel underwater visual localization methods. The Intel computer-vision library OpenCV was used at a speed of 3 frames per second. 40 st / July 2013 This threshold is therefore also dependent on the sonar attack angle. With a small attack angle, it becomes more diffcult to distinguish obstacles from a tank wall. The following standard deviation is also used in the later construction of an extended Kalman flter (EKF) SLAM algorithm: Obstacle Detection. Given the vehicle pose: it is easy to get range estimation: between the sonar and its attack or refection point on the tank wall surface. If the range measurement: satisfes: then the attack point is considered as an obstacle point. Each obstacle block is modeled as a pair of its start and end points. www.sea-technology.com

Articles in this issue

Links on this page

Archives of this issue

view archives of Sea Technology - JUL 2013
loading...
Sea Technology
Welcome!
If you're not a subscriber, please click here.