On a table top, there are a set of different objects lying around. The goal is to find a plan according to which the robot can put the objects to some predefined final pose. The final and initial poses of the objects have overlapping in the work space, which makes the problem more challenging than simple pick and place.
The task and motion planning algorithm generates a feasible and reasonable sequence of action to manipulate the objects on a table top. The algorithm runs successfully in simulation as the following video shows:
Here is the one example of the implementation on real robot in a ROS environment. In this video, three objects (cup, bowl and a coffee box) have been considered.