Artificial Intelligence


In the 1980s, Marc Raibert and his team at the M.I.T. university has introduced legged robots which are able to walk over uneven tarrain. The assumption was, that a single leg hopper has to jump continuously and that a software controller should stabilize the system. But before a robot is able to walk over ground, the more natural motion is to climb in the trees. Apes for example, aren't familiar with walking with feets but they are preferring arms in connection with the branches of a tree.

In the simple case a brachiating robot contains of 2 dof.[1] Both end effectors are able to close the gripper around a peg. If the control actions are given at the right time, the system is able to move forward by using subskills. These skills are ordered in a pattern, similar to a gait pattern, but not at the ground in favor in the air.

A more complex climbing robot contains of more than 2 dof.[2] Similar to the first example, the gripping action is divided into a sequence of subactions which is planned by a grip sequencer. The overall trajectory is determined by the path planner.


  1. YAMAKAWA, Yuji. "Brachiation motion by a 2-DOF brachiating robot with hook-shaped end effectors." Mechanical Engineering Letters 4 (2018): 18-00094.
  2. Zhu, Haifei, et al. "Transition analysis and its application to global path determination for a biped climbing robot." Applied Sciences 8.1 (2018): 122.