Position-based visual servoing via model-based design
Loading...
Date
2019-05
Authors
Choo, Jie Liang
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Visual servoing is a technique where the motion of robot is controlled by the
computer vision. There are two common visual servoing techniques that are currently
been used in the world, which are Image-Based Visual Servoing (IBVS) and PositionBased Visual Servoing (PBVS). Various approaches such as Component-Based
Design (CBD) and Model-Based Design (MBD) are used to design the mentioned
visual servoing. Therefore, in this project, a combination of MBD with PBVS is
proposed to design a follower robot, which is programmed to sense and track leader
robot in 3D manner. In DoE, the follower robot is designed with an image sensor and
ultrasonic sensor, allowing it to extract 3D data such as height, width of the pattern
and the distance between leader and follower. A set of different scenarios are
introduced to evaluate the performance of the follower robot, which is also categorized
into straight line (Trajectory 1) and curved (Trajectory 2) trajectories. Experimental
results show that the follower can track the leader easily for Trajectory 1- straight line,
with high stability and short time, such as 1.40s for Experiment 1.1, 1.47s for
Experiment 1.2 and 0.6s for Experiment 1.3. However, for Trajectory 2, the time
required for overall tracking is slightly increased, such as C, thus lowering the
efficiency of the tracking. In conclusion, the model-based design can achieve the
minimal requirement of PBVS, which is to track the leader. The method however may
be improved further in the future to minimise the overshoot and the optimise the
tracking time.