Safe Cooperation between Human Operators and Visually Controlled Industrial Manipulators

Human-Robot Interaction 204 with its environment.Nevertheless, the robot needs sensors in order to adapt its behaviour depending on the changes in the environment when human-robot interaction tasks are performed.In this work, the robot has a camera mounted at its end-effector to control its movements according to a visual servoing method.The main objective of visual servoing is to minimize the error between the image obtained at the first pose of the robot and the image obtained at the target position of the robot.Visual servoing is adequate to control the robot in situations where external or not planned objects enter the robot workspace, avoiding a possible collision.When the robot has to perform a planned path in a 3D space limited by other objects, classic image based visual servoing fails to track this planned path.Previous research on visual path tracking has tried to solve this problem by making use of visual servoing to follow a desired image path previously sampled in time.These systems can be modified in order to obtain a human safety algorithm.In this way, when a human is dangerously close to the robot, the path tracking must be stopped.However, after the danger of collision has disappeared, previous image trajectory tracking methods based on visual servoing fail to return to the initial path because they are time-dependent.Therefore, a time-independent behaviour of the control system is crucial to develop interactions with the workspace.This time-independent behaviour makes the robot continue on the same path from the same point that it was tracking before the detection of the human.The method described in this chapter guarantees the correct tracking in the 3D space at a constant desired velocity.As shown before, a safety behaviour which stops the normal path tracking of the robot is performed when the robot and the human are too close.This safety behaviour has been implemented through a multi-threaded software architecture in order to share information between both systems.Thereby, the localization measurements obtained by the human tracking system are processed by the robot control system to compute the minimum humanrobot distance and determine if the safety behaviour must be activated.This chapter is organized as follows.Section 2 describes the human tracking system which is used to localize precisely all the limbs of the human operator who collaborates with the robotic manipulator.Section 3 presents an introduction to visual servoing and shows how the robot is controlled by a time-independent path tracker based on it.Section 4 describes how the safety behaviour which avoids any collision between the human and the robot is implemented.This section presents the hierarchy of bounding volumes which is used to compute the human-robot distance and modify the movements of the robot accordingly.Section 5 enumerates the results obtained in the application of all these techniques in a real human-robot interaction task where a fridge is disassembled.Finally, the last section presents the conclusions of the developed research. Human tracking system Components of the human tracking systemThe human operator who interacts with robotic manipulators has to be localized precisely in the industrial workplace because of two main reasons.On the one hand, the knowledge of the human location enables the development of safety behaviours which avoid any risk for the physical integrity of the human while their interaction takes place.On the other hand, the localization of the human operator can also be taken into account to modify the movements of the robot accordingly.The movements of the robot are adapted to the human's behaviour and thus more flexible human-robot interaction tasks can be implemented.www.intechopen.