Accurate Step Length Control Strategies for Underactuated and Realistic Series Elastic Actuated Hoppers via High Order PFL
- Author(s): Terry, Patrick W.
- Advisor(s): Byl, Katie
- et al.
Among the different types of legged robots, hopping robots, aka hoppers, can be classified as one of the simplest sufficient models that capture the important features encompassed in dynamic locomotion: underactuation, compliance, and hybrid features. There is an abundance of work regarding the implementation of highly simplified hopper models, the prevalent example being the spring loaded inverted pendulum (SLIP) model, with the hopes of extracting fundamental control ideas for running and hopping robots. However, real world systems cannot be fully described by such simple models, as real actuators have their own dynamics including additional inertia and non-linear frictional losses. Additionally, implementing feedback control for hopping systems with significant amounts of compliance is difficult as the input variable does not instantaneously change the leg length acceleration. The current state-of-the-art of step length control in the presence of non-steady state motions required for foothold placement is not precise enough for operation in the real world. Therefore, an important step towards demonstrating high controllability and robustness to real-world elements is in providing accurate higher order models of real-world hopper dynamics, along with compatible control strategies.
Our modeling work is based on a series-elastic actuated (SEA) hopping robot prototype constructed by our lab group, and we provide verifying hardware results that high order partial feedback linearization (HOPFL) can be implemented directly on the leg state of the robot. Using HOPFL, we investigate two paths of compatible trajectory generation that can accomplish desirable tasks such as precise foothold planning. We investigate the practicality of using SLIP-based trajectory generation techniques on more realistic hopping robots, and show that by implementing HOPFL directly on the robot's leg, we can make use of computationally fast SLIP-based approximations, account for non-trivial pitch dynamics, and improve the state-of-the-art of precision step length control for SEA hoppers. We also consider control strategies towards hoppers for which SLIP-based trajectories may not be compatible, by planning all ground reaction force vector (GRF) components during the stance phase concurrently, using a lower order and very general model to construct trajectories for the system's center of mass (CoM), and maintain body stability by controlling the orientation of the GRF directly. While not purely analytical as our SLIP-based approaches, this method is general enough to work on a variety of hopping robots that are not necessarily kinematically structured resembling the classical SLIP model.