Vision System and Calibration

This monograph highlights how different technologies, namely, vision, robotic manipulation, sensitivity and uncertainty analysis, kinematic and dynamic identification, force control etc., can be combined to get an end-to-end fully functional system. This chapter acts as a window to this treatment to explain perception of the external environment using sensors. The environment to be sensed contains a large number of symmetrical and identical feature-less, texture-less objects (black cylindrical pellets) randomly piled up in the bin, with arbitrary orientations and heavy occlusion. The task at hand is to use a set of sensors with complementary properties (a camera and a range sensor) for pose estimation in heavy occlusion, accordingly orienting the manipulator gripper to pick up a suitable pellet, and this process is repeated to pick all the pellets one-by-one. Furthermore, the manipulator avoids collision with the bin walls by identifying and characterizing the cases when the object is present in a blind spot for the manipulator. Thus, this chapter lays the foundation for the subsequent chapter, i.e., uncertainty and sensitivity analysis of the vision-based system. These chapters also prove that building a fully functional robust pipeline in real-world scenarios is quite challenging. Particular aspects that are of paramount importance for the successful accomplishment of the task is also presented. Another critical aspect to emphasize here is that it deals with real-world industrial problems, i.e., bin-picking. Different approaches that work best due to variations in assumptions and experimental protocols, e.g., sensors, lighting, robot arms, grippers, and objects, are dealt with in detail. A layout of the literature survey is provided as visual chart to give the major technologies and components involved in this chapter. The contents of this chapter will be suitable for practitioners, researchers, or even novices in robotics to gain insight into real-world problems.