UCLA

Robots have become more ubiquitous in daily life than ever before. Package delivery, factory automation, and transportation are just a few of the areas that have started to see robots take defining roles. Despite this, robotics still has a long way to before realizing the end goal of living and working amongst humans. The real world presents many challenges including dynamic obstacles, uneven contact surfaces, and changing operating conditions that require robots to quickly adapt on the fly in order to be successful. While current state of the art robots have become quite useful as single task platforms, there is still a need for a more generalizeable solution in order to meet the challenging needs of real world tasks. In order to truly adapt to an environment made for a human, robots must now take on the same humanoid form.

To this end, a bipedal locomotion control stack was developed for the torque controlled humanoid Artemis. A model based hierarchical stack using a linear inverted pendulum inspired footstep planner, polynomial swing leg trajectories, and an implicitly weighted hierarchical whole body controller were tested to yield stable walking at a maximum walking speed of 1.2 m/s. In addition, to verify the effectiveness of the platform and its underlying software under real world conditions, an autonmous soccer playing software stack was implemented to compete in Robocup 2024. The stack includes a vision model capable of detecting balls and field landmarks, a localization algorithm to estimate the game state, a behavior tree to make real time soccer decisions, and a path planner to avoid obstacles through gameplay. Furtheremore, a soccer kick was developed that was capable of kicking the soccer over half of the miniturized field in a single kick. These advancements resulted in a first place finish at Robocup 2024.