There are various ways in which we can perform pick and place tasks. For example, we could send the robot a predefined sequence of joint values directly, which will cause the robot to always perform the same predefined motion. In this case, we must always place the object in the same position. This method is called forward kinematics, because we have to know in advance the sequence of joint values,   in order to perform a certain trajectory.

Another method would be by using inverse kinematics (IK), without any vision feedback. In this case, we provide the robot with the pose (x, y, or z) where the object to pick is, and by doing some IK calculations, the robot will know which motions it needs to perform in order to reach the object's pose.

Similarly, another method would be to use inverse kinematics with vision support or feedback. In this case, we would use a node that would identify the pose of the object by reading the data from the sensors of the robot, such as a Kinect camera. This vision node would provide us with the pose of the object, and again, by using IK calculations, the robot would know the required motions to perform in order to reach the object.

We are probably thinking that the most efficient and stylish method is the third one, and we're right. But, this method will need object recognition knowledge, which is not the subject of this course. So, for this course, we are going to use the second method, which in the end is the same as the third one, but without the object recognition part. If you already have some perception knowledge, you can freely add an object recognition node that sends the pose of the object without having to set this pose in the code itself.

So, to summarize, what we are going to do in the next subsection of this chapter is create a node that will send an object pose to the Grasping Server, in order to receive the appropriate joint values to reach that object.