Study On The Multi-robot Formation In Obstacle Environments

Multi-robot formation system refers to a group of robots which move together with a certain formation. It is important in research of multi-robot system, and can be applied in many fields like aviation, aerospace, transportation, military, and so on. So far, many typical formation methods have already been proposed, including leader-follower, virtual structure, behaviorism, etc. These methods can guarantee the convergence of the whole formation. However, what they mainly focus on is the inner problem, but not enough for the outer ones. Thus, they can't be used in living environment in which there exist lots of obstacles both static and dynamic. As a result, it narrows the application of multi-robot formation system in the field of service robots. So, here we divide the whole formation system into two parts - trajectory and formation shape, and realize its safety and formation maintenance in obstacle-environment through path-planning and behavior-switching. Details are as follows:1. A 3-level scheme of the formation control is discussed, including Task Level, Planning Level and Control Level. The formation maintenance based on leader-follower method and movement-control based on artificial potential method is also analysised in details.2. To deal with the contradiction between formation maintenance and obstacle avoidance, a method of rigid formation control in environment with static obstacles in it is propose. It regards the robots in formation as an entirety by fusing information of sensors of every individual robot and designing a particular angular velocity. And this has already been tested and verified by experiments.3. As common dynamic obstacles exist in living environment are usually autonomous, a method of rigid formation control based on FSM (Finite State Machine) is proposed. On the basis of recognition of dynamic and static obstacles, it avoids the obstacles successfully by changing the velocity of formation. It can guarantee the integrity of the formation while avoiding the obstacles, and is easy to realize as well. This has already been tested and verified by experiments. 4. In order to clear the accumulated error taken by the robot odometer, considering cameras usually exist in obstacle environments, we study how to locate obstacles by a single camera. This thesis proposed a method of locating the robot center based on the RGB-analysis and morphology with a kind of appropriate marks. It has been tested and verified by experiments to have enough accuracy and can be used in different backgrounds, especially where there is reflective.The validity of all mentioned above has been tested and verified by simulations and experiments, and they have the potentials and are valuable for more complicated environments.