Trajectory optimization is an important means to improve performance i
n many industrial and scientific processes such as robotics or aerospa
ce applications. A combination of SQP methods with direct collocation
or multiple shooting discretizations has proven very successful for su
ch problems. In realistic applications, however, very large QP subprob
lems with thousands of constraints may arise. The paper presents an ap
proach based on highly efficient solution of the block-sparse QP by a
structured interior point method.