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/679071
36 st / May 2016 www.sea-technology.com and isolate the data from scan results. We employ a median flter to clear noise in order to improve algorithm effciency. The median flter utilizes mid-value in the neighborhood of a point to represent the point, and it can eliminate isolated noise. The mid-value of a sample is selected in the scanning area of the lidar after ar- ranging the data from large to small in the scanning scope. Selecting the mid-value utilizes a 2D sliding formwork to arrange a monotone 2D array of data. Obstacle Avoidance Algorithm The USV is confronted with the nav- igation problem of avoiding unknown obstacles quickly. The environment and the direction of motion should be taken into account. We divide the measurement range of lidar into two layers: accurate obstacle avoidance and fuzzy obstacle avoidance. Setting a secondary target prevents the robot from leaving a set region. The fuzzy algorithm is carried out with the larger measuring range of li- dar instead of the accurate algorithm. The corresponding right-front (RF) an- gles are -5° and 60°. The correspond- ing front (F) angles are 60° and 120°. The corresponding left-front (LF) angles are 120° and 185°. The output variable UD denotes the direction of the USV, and its corresponding angles are -5° and 185°. Three obstacles were set up. Experiments In order to verify the reliability of the algorithm, simulation was done by MATLAB. The path of the modi- fed algorithm proved to be smoother than the path of the conventional al- gorithm. When the robot encountered obstacles, the modifed APF algorithm controlled the robot to move along the obstacle boundary to avoid them. The path of the modifed algorithm is shorter than the path of the conven- tional algorithm. For the conventional APF, the path length was 150 m. For the integrated algorithm, the path length was 123 m. A second target was added to cre- ate a trajectory for the USV. When the conventional algorithm is adopted, the modifed algorithm utilizes the method of setting a secondary target to escape from the local minimum area. Af- ter reaching the secondary target, the robot moves to another target. The simulation result shows that the modifed APF is feasible. We decided to employ the modi- fed APF algorithm in the accurate layer, so that the USV will move to a target according to fuzzy rules. Conclusion We employed a novel, inte- grated algorithm for USV obstacle avoidance consisting of accurate and fuzzy rules to deal with the bugs of the traditional APF algo- rithm and sensor errors. The USV is controlled by the modifed APF algorithm in the effective mea- suring range of lidar sensors. If it drops into the local minimum area, the algorithm will drive the USV to search for a secondary target until it leaves the area. The simulation shows that the proposed algorithm can con- trol the USV to avoid obstacles and fnd an appropriate path toward a tar- get compared with the conventional method. However, we only used a sample method for obstacle detection, and we've done little work on dynamic object tracking. In the future, we will continue to improve our obstacle avoidance algorithm and put more emphasis on obstacle detection and object tracking based on lidar data. ST Peng Wu received his master's degree in petro- leum engineering from Changzhou University, China in 2013. He is cur- rently a Ph.D. student in the Department of Mechatronic Engineer- ing, Shanghai University, China. His research interests are USV navigation, trajectory tracking and mechanical design. Dr. Shaorong Xie re- ceived B.S. and M.S. degrees in mechanical engineering from Tianjin Polytechnic University in 1995 and 1998, respec- tively, and a Ph.D. degree in mechanical engineer- ing from the Institute of Intelligent Machines at Tianjin University and the Institute of Robotics and Automatic Information Systems, Nankai University in 2001. Her research areas include advanced robotics technologies and image monitoring systems. Hengli Liu is currently a Ph.D. candidate in the Department of Me- chatronic Engineering, Shanghai University. Obstacle detection result.