Controller for stair ascent and descent for a humanoid robot using bio-inspired mechanisms

Being able to increase humanoid robots locomotor capabilities is a great challenge. Making them able to deal with some known perturbations of the ground will widen their range of possible actions in our highly unpredictable human-made environment. With a wider working environment, it might be expected that humanoid robots could help humans in a lot of new situations. In this work, we design a bio-inspired controller for the humanoid robot CoMan, to enable it to ascent and descent stairs in simulation. The controller has an existing bio-inspired controller for flat ground as starting point and uses mechanisms such as a musculoskeletal model, reflex rules and a central pattern generator. It is developed in a two-dimensional environment and tested using the simulation software Robotran. We enable the robot to ascent and descent stairs of a rise of up to 6 cm (14 cm for humans given the size of the robot), which corresponds almost to a standard size of rise for humans. The minimization of the metabolic energy consumption is privileged over the control of the length of the steps. We test also the robustness of the controller during ascent and descent. Moreover, the robot is also able to perform the transition between flat ground and ascending stairs using a specific controller during transition. The transition between ascending stairs and flat ground is also possible. Furthermore, we identify the most important muscles of the model during locomotion and the most important parameters of the controller that might be used later to modify the gait of the robot. Finally, we compare the muscles activations and the kinematics of the robot with human data.