Finding Locally Optimal, Collision-Free Trajectories with Sequential Convex Optimization

Experiments in Reality

In the following videos we show our optimization algorithm planning trajectories of various degrees of freedom (DOF) and then the execution of these trajectories on the Willow Garage PR2.

Picking up a crumpled piece of paper on the table and throwing it to the recycle bin under the table
Opening a door to grab a towel inside a room
Navigating through an office with various obstacles in the way

Experiments in simulation

Humanoid walking

Box picking

The box is subject to a partial orientation constraint, i.e., "this side up"
The robot must grasp the boxes and bring them to a drop-off location to its right.