Control and Gait generation of a 3D biped robot in Gazebo

Today’s bipedal robots still are behind humans in terms of efficiency, speed, and resilience in movement. Consequently, this thesis proposes a control strategy for dynamic walking inspired by human motion control principles. Central to this approach are the utilization of passive dynamics, hierarchical control mechanisms, and reflex actions, all achieved without necessitating a complete dynamic model. Walking stability is ensured through a series of postural reflexes linked to the extrapolated center of mass motion. It is demonstrated that only a minimal set of joints need simultaneous active control throughout different walking phases. Alongside detailing the control strategy, the thesis introduces an anthropomorphic biped model and highlights the importance of features such as compliant actuation for optimal walking performance. Specifically, the proposed method relies on joints with characteristics akin to the human muscle-tendon system, including non-self-locking, torque-controllable mechanisms with parallel elasticity and minimal friction. The effectiveness of the approach is validated through simulations of 3D dynamic walking, revealing an efficient, smooth, and rapid gait capable of handling significant disturbances. Moreover, the resulting joint trajectories closely resemble human walking patterns. In recent years, numerous researches have been done based on simulation of legged mechanism,
especially on biped robots simulation and control. This research focuses on the biped robot simulation in Gazebo and the control of it over various trains such as flat environment, ascending and descending surfaces, passing a ditch and different obstacles with the aid of 3D modeling methods. the bipedal robot‘s movement is controlled by 3 different methods (Neural Network controller, Fuzzy Logic Control and PID controller). Finally the results of all these controllers are provided and the compression is done in later chapters.