Control and Design for a Multibody Model of Biped-Flexible Wheeled Robot

A variety of robots inspired by humans have been introduced in recent years. Many of them are conceived as structurally flexible and with elastic parts. Deformable components can be smaller than rigid ones, resulting in a lightening of the structure. They can also store energy and release it at a later time resulting in a more energy-efficient solution. Biped robots can walk not only on a flat floor but also on uneven environments and stairs among challenging situations. However, walking is not an easy way to move because it requires many joints and complicated control structures. On the other hand, wheeled robots are widely used for flat terrain because of high stability, high efficiency, and simple controller, nevertheless, they suffer from several limitations that reduce their application. Wheel-legged robots retain the advantages of both wheeled and legged robots. Many biped robots use wheels or flexible elements, but rarely a combination of the two. In this thesis, an innovative type of biped robot named Rollo is analyzed by simulation. Rollo is an innovative type of small biped robot with wheels developed by the BrainHuRo project which was aimed at conceiving humanoids controlled by the human brain for people affected by the Amyotrophic Lateral Sclerosis disease. The innovative parts of the robot are the use of the cylindrical helical springs in the leg with the wheeled feet. This flexible part allows the robot to perform an alternate biped walking without the use of joints and motors for the knees and ankles but with only two motors in the feet. A rigid 3-D model of the Rollo robot is realized in Solidworks and then transferred to Matlab/Simulink. Two different configurations are considered, one with 2 springs in each leg the other one is 3 springs in each leg. The Simulink multibody model integrates also the flexible links and the control part. The controller of each foot is a PID that computes the wheel torque starting from the position error of the wheel compared to the reference signal. Different reference paths are considered, in particular a linear motion, a walking motion, and a rotation in place. Each of them has in 4 different velocities from 0,025 m/s to 0,2 m/s. Two accelerometers (at the ankle and the hip of the robot) measure the vibrations of the robot along the 3 axes. The obtained data are analyzed and compared with the ones coming from the real prototype.