Better Stair Climbing By Using Symmetry

Roboticists often approach problems of control from a brute force modeling perspective. Instead, we construct a solution based on understanding of the problem’s symmetries. We are attempting such a solution to the problem of robust stair-climbing. For a robot executing a slow periodic gait - one that builds up no momentum - the state of the robot’s limbs is captured in one cyclic phase variable. A second “shift” variable, captures the robot’s location relative to the stairs’ periodic pattern. A third “skew” variable, captures the heading angle of the robot relative to the stairs. When climbing an infinite staircase, these three variables complete the robot’s state up to symmetry.

We designed an open-loop climbing gait cycle for Argus, a low-cost meter-long hexapedal robot made of foam-core. Using data from six inexpensive leg-mounted IR sensors, we could predict failure of the open-loop gait. We showed that motor loads and IR sensor readings suffice to measure “shift” and “skew”. We hope to use these measurements to provide feedback that will make our climbing gait robust to variations between stairs and mishaps that occur during stair climbing.