Researchers at the UC Berkeley Hybrid Robotic Group and CMU are working hard to make sure their robots do not tiptoe into rugged terrain . Through machine learning and ATRIAS robots, teams can "teach" robots to walk through stepping stones that they have never seen before.
Their robots, described here, are unique in that they are bipedal and use a mixture of balance and jump to make sure not to rock the blocks.
"What is different in our methods is that they allow dynamic walking as opposed to slower quasi-static movements that robots tend to use," the researchers write. "By reasoning about the nonlinearities of the dynamics of the system and taking advantage of recent advances in optimal and non-linear control technology, we can specify the control objectives and desired behaviors of the robot in a simple and compact form. This means that our robots can walk on unobtrusive terrain without slipping or falling, supported by careful maths and beautiful experimental videos. "
The robots are currently "blind" and can not use visual input to plan their next move. However, with a robot called CASSIE, they will be able to see and feel the stones as they go, ensuring that they do not rock into the fire of action … or battle .