|M.Sc Student||Meital Shaked|
|Subject||Combination of Coarse and Fine Visual Servoing Control for|
a Polar Precision Sprayer
|Department||Department of Civil and Environmental Engineering||Supervisor||Professor Linker Raphael|
|Full Thesis text - in Hebrew|
The use of mobile autonomous robotic systems in dynamic and versatile environments has increased in recent years. Therefore, there has been a growing interest in the need for sensor fusion in robotic systems to navigate the robot within the environment while interacting with it.
Visual feedback is a prevalent approach for this purpose, due to the fact that it mimics the human sense of vision. Visual information collected from the environment during the performance of a task is essential for robots to execute flexible and autonomous tasks. The fusion between vision and robots, using closed loop position control, is referred as Visual Servoing. The control task for these systems is defined as end effector pose control relative to a specific target object.
In the present study, a visual servoing system was built and tested. The system includes a 2-DOF (rotational and axial) polar arm consist of rotating arm and a mobile cart attached to it (integrated with 3 cameras). The system is to be integrated in the future in the field of precision agriculture, and with adequate modifications, is to be part of a precision sprayer system.
The main goal of the research was the development of a visual servoing system for controlling the pose of a robotic arm using an integrated control system combining coarse and fine control loops. Each control loop includes a camera and image processing unit. The system was designed as a multilayer control system. Its architecture includes combining data from different sources: layers of control (coarse and fine), a layer of path planning, a layer of execution motion, etc.
A variety of tests were performed on the system in order to evaluate its performance. Test results have shown the efficiency of the fine control loop: significant location errors after motion completion according to the coarse control were decreased to small location errors after completing the motion according to the fine control.
In addition, tests results showed that the initial position of the end effector or the target location in the workspace has no influence on the final accuracy of the system.