Mobile Robot Visual Servo

Mobile robot is a comprehensive system which includes the environment sensing, dynamic decision-making, action control and execution. Vision is an important means of observing and understanding the world, about 80 percent human information obtained from the outside world is obtained by vision. With the development of the information technology, computers,robots and other intelligent machines are given human visual cognitive function, which make them have the cognitive ability about the real world through two-dimensional image. Robot visual servoing is an important field of robot research, and is also an important aspect of robot navigation. Visual servoing improves the performance of the existing intelligent robots and make robots be more widely used in the military, space, industrial and other fields.Robot visual servoing has three types:Position-based visual servoing (PBVS),Image-based visual servoing(IBVS) and 2-1/2-D visual servoing. Because the PBVS needs a priori 3-D knowledge of the scene, and 2-1/2-D visual servoing requires perfect knowledge of the camera calibration parameters and an estimation of the camera position, Image-based visual servoing becomes the focus of this paper.In IBVS, the usual method is an approximate input-output linearizing feedback control method. The IBVS method need some approximate treatment and an estimation of the distance between the initial position and desired position. All of these will give the result some error, especially when the error of the distance estimation is huge, which is likely to cause the task failure. So this paper presents a new image-based visual servoing method. The method combing epipolar geometry with triangle geometry can avoid approximate treatment and does not need an estimation of distance between the initial position and desired position, so it can complete the task well. In IBVS, it is difficult to ensure that the image features has been in a camera domain when using pinhole camera. If the feature points are not captured by the camera is likely to lead directly to the failure of the mission. Panoramic camera can solve the problem of restricted vision domain, so this paper also studies the visual servoing algorithm with panoramic camera.In the end, the simulation results compare the existing algorithm and the algorithm, and prove the effectiveness and feasibility of this algorithm.