SIA OpenIR  > 海洋信息技术装备中心
多海洋机器人协同避碰行为研究
Alternative TitleResearch on Collaborative Collision Avoidance of Multiple Marine Robots
吴函
Department海洋信息技术装备中心
Thesis Advisor徐红丽
Keyword海洋机器人 局部感知 协同避碰 多机器人
Pages80页
Degree Discipline控制理论与控制工程
Degree Name硕士
2021-05-21
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract在海洋机器人应用时,常以水深作为其中一项重要指标,并将水下0-5米的范围称为近水面区域,近水面区域作为水下与水上交接的重要区域,无论是出入深水时对机器人进行位置调整,还是对水面物体进行监测,都需要在近水面区域进行作业,而近水面的机器人应用尚存在诸多问题需要解决,例如近水面强波浪干扰降低了机器人的控制精度、浪涌引发的自由液面水动力问题、水面波光造成图像亮度不均衡问题、水下的物体会出现倒影的问题、近水面区域水声不易通信问题、障碍物影响机器人航行问题等,由于其中障碍问题最为重要。发生碰撞极易造成设备损坏,无法完成预定任务的同时还造成大量财产损失,因此需要优先解决其中影响障碍物感知的倒影问题以及面对不同障碍时如何安全进行避碰的控制算法问题。本文的多海洋机器人协同避碰行为主要是基于视觉方式的局部感知与局部通信展开研究。不使用水声方式感知障碍或互相通信有以下几点原因:第一水声在近水面的表现不佳,一般水声通信要求机器人在较深的位置进行作业,或是在一端水声设备放置于近水面,另一端水声设备放置于较深的位置时才能保持较好的通信效果。第二水声在海洋中无法保持其隐蔽性,在很多海洋应用场景中,都要求机器人不能使用水声设备,防止信息泄露。第三在无中央控制器的群集系统中,水声较难满足大规模机器人的通信需求。因此本文基于视觉方式进行局部感知与通信更符合这种近水面应用场景,并有很大的发展与应用空间。对于上述情况下的多海洋机器人群集系统,分布式决策是这种系统的主要特征,因此,这种群集系统的协同避碰控制可以先对每个机器人的运动模型进行假设,然后根据个体的运动模型和障碍的大小不同去设计具体的群体协同避障控制方法。在这种协同避碰方法中每个机器人都是根据相邻机器人的行为进行相应的动作,使未检测到障碍的机器人也能跟随其他检测到障碍的机器人一同完成避障,从而使群体协同避碰算法的避障效率达到“1+1>2”的效果。具体研究内容如下:(1)视觉感知障碍。机器人在近水面需要避碰的对象主要分两类,一类是海面上的漂浮物,另一类是在近水面区域航行的同类型机器人。因此这一部分研究内容将使用YOLOv3算法对两类不同的避碰对象进行识别,然后针对倒影影响识别精度的问题,提出一种平滑化区域填充方法提高了视觉方式识别机器人的准确率。(2)类似浮球或者机器人这种小型障碍情况下,基于局部感知的多机器人协同避碰方法。在协同避碰方法中,所有机器人都不具备全局感知障碍物的能力,只有在障碍物附近时,才能使用局部感知能力获得障碍信息,由于机器人群避开小型障碍的时候局部感知依旧能使群内个体保持连通,因此可以为所有机器人设置目标点引导机器人群的整体航向,然后综合考虑自己航向和目标点位置还有障碍的位置分情况调整智能体的避障行为,解决了多个障碍影响之下可能无法避开障碍的问题。(3)类似大型船这种大型障碍情况下,基于局部感知的多机器人分群避碰方法。若有大型船只或类似大型障碍物停留在航线中时,整个机器人群很可能会因为避碰行为发生分裂,分裂之后的子群规模差距悬殊,局部感知方式无法连通两个分裂后的子群,子群中机器人数量太少也无法继续执行任务,针对这种情况,本文提出一种融合拓扑交互与信息耦合方式的分群避障方法,不仅提高了机器人群集组群和分群的速度,还提升群集等规模分裂的几率。
Other AbstractIn Marine robot applications, often in the depth of the water as one of the important indicators, and the range of 0 to 5 meters underwater called near surface area, near surface area as an important area of underwater and water transfer, both in and out of the deep position of robot carried out, or on the surface of the equipment for testing, need to be in the near surface area to carry on the task, And near surface of the robot application still exist many problems to be solved, such as strong near surface wave interference to reduce the control precision of the robot, the surge caused by free surface hydrodynamic problem, the water sparkling image brightness uneven problem, there will be a reflection of underwater object, near surface area underwater acoustic communication problems and obstacles robot navigation, etc., Because the problem of obstacles is the most important. Collisions are easy to cause equipment damage, which can not complete the predetermined task but also cause a large number of property losses. Therefore, it is necessary to prioritize the reflection problem affecting the perception of obstacles and the control algorithm problem of how to safely avoid collisions in the face of different obstacles. The cooperative collision avoidance behavior of multi-ocean robots in this paper is mainly studied based on the local perception and local communication in the visual mode. Do not use water way feel disorder or communicate with each other for the following reasons: the first underwater acoustic underperforming in near surface, general underwater acoustic communication requires robots in the deep position assignments, or placed at one end sonar equipment in near surface, the other end of the acoustic equipment placed in a deeper position to maintain good communication effect. Second, underwater sound cannot be concealed in the ocean. In many ocean application scenarios, robots are required not to use underwater acoustic equipment to prevent information leakage. Third, in the cluster system without central controller, underwater acoustic is difficult to meet the communication requirements of large-scale robots. Therefore, the local perception and communication based on visual method in this paper is more in line with such near-surface application scenarios, and has a great space for development and application. For the cases of Marine robot cluster system, distributed decision-making is the main characteristic of this system, therefore, the cluster system on the synergy collision can control for each robot motion model is assumption, and then according to the size of the individual motion model and obstacles to design different specific group collaborative obstacle avoidance control method. In this collaborative collision avoidance method, each robot takes corresponding actions according to the behavior of its neighbors, so that the robot without detecting the obstacle can follow other robots to complete the obstacle avoidance together, so that the obstacle avoidance efficiency of the group collaborative collision avoidance algorithm can reach "1+1 >2 "effect. Specific research contents are as follows: (1) Obstacles are perceived visually. The objects that robots need to avoid collision in the near water are mainly divided into two categories, one is floating objects on the sea, and the other is the same type of robots sailing in the near water area. Therefore, this part of the research will use YOLOV3 algorithm to identify two different types of collision avoidance objects. Then, aiming at the problem that reflection affects the recognition accuracy, a smoothing area filling method is proposed to improve the accuracy of robot recognition by visual method. (2) Collaborative collision avoidance method for multiple robots based on local perception in the case of small obstacles such as floating ball or robot. In a coordination collision avoidance method, all robots don't have the ability of global sense obstacles, only near the obstacles is to use local awareness for obstacle information, due to the machine to avoid the small crowd barriers when local perception can still remain connected within the group of individuals, and so can set a target for all robot guide machine crowd the whole course, Then the obstacle avoidance behavior of the agent is adjusted according to its own heading, target point position and the position of obstacles comprehensively, so as to solve the problem that it may be impossible to avoid obstacles under the influence of multiple obstacles. (3) Collision avoidance method for multi-robot clustering based on local perception under large obstacles such as large ships. If there are large ships or similar large obstacles to stay in line, the whole machine group is likely to occur because of the collision avoidance behavior split, split after the subgroup size gaps, divides local perception mode cannot connect the two subgroups, subgroup in the robot cannot continue to perform a task in number, for this kind of situation, In this paper, a clustering obstacle avoidance method combining topological interaction and information coupling is proposed, which not only improves the speed of robot clustering and clustering, but also improves the probability of clustering and other scale splitting.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/28946
Collection海洋信息技术装备中心
Affiliation中国科学院沈阳自动化研究所
Recommended Citation
GB/T 7714
吴函. 多海洋机器人协同避碰行为研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2021.
Files in This Item:
File Name/Size DocType Version Access License
多海洋机器人协同避碰行为研究.pdf(5253KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[吴函]'s Articles
Baidu academic
Similar articles in Baidu academic
[吴函]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[吴函]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.