Planner-aided Design of Ladder Climbing Capabilities for a DARPA Robotics Challenge Humanoid

This paper describes preliminary steps toward providing humanoid robots with ladder climbing capabilities. Although seemingly straightforward for humans, ladder climbing is quite challenging for humanoid robots due to differences from human kinematics, significant physical stresses, simultaneous coordination of four limbs in contact, and limited motor torques. We present a planning strategy for the the Hubo-II+ humanoid that automatically optimizes motion primitives offline and adapts them on-line to novel ladders. The adapted motions are guaranteed to satisfy contact, collision, quasi-static balance, and torque limit constraints. This method is used to automatically test a number of strategies, including forward and backward climbing, on a variety of ladders in simulation. This planner-aided design paradigm has proven useful for hardware design because it permits the rapid testing, verification, and optimization of hardware changes by automatically generating novel climbing strategies.