Control design for collaborative object transportation with differential drive robots : Enforcing distance and angle constraints via Control Barrier Function

Abstract

Collaborative object transportation has been done in the past with a variety of methods and robots, but not all methods have a mathematical proof of safety. In safety-critical applications, such a guarantee is one of the main design parameters. This thesis proposes controllers that use Quadratic Programming - Control Barrier Function (QP-CBF) to provide this certainty.

The controllers guide a pair of differential drive robots towards a target pose across an unknown environment with obstacles. The object transport is modeled as a constraint in the distance between the robots and another limit on the angle between each robot and the object carried. Avoiding other robot groups is also considered. Pose control is handled with CBF as well.

Simulations show that the controllers are successful at leading the robot group to the target pose, keeping them inside the safe set of states. The centralized controller performs better than the polar coordinates pose controller with similar CBFs, while the distributed controller takes longer to reach the target. An experiment with a couple of TurtleBot3 robots was done as well. There, a slightly modified centralized controller managed to guide the robots to the target pose while avoiding an obstacle.