Toward a Data-Driven Template Model for Quadrupedal Locomotion

Abstract

This work investigates a data-driven template model for trajectory planning of dynamic quadrupedal robots. Many state-of-the-art approaches involve using a reduced-order model, primarily due to computational tractability. The spirit of the trajectory planning approach in this work draws on recent advancements in the area of behavioral systems theory. Here, we aim to capitalize on the knowledge of well-known template models to construct a data-driven model, enabling us to obtain an information rich reduced-order model. In particular, this work considers input-output states similar to that of the single rigid body model and proceeds to develop a data-driven representation of the system, which is then used in a predictive control framework to plan a trajectory for quadrupeds. The optimal trajectory is passed to a low-level and nonlinear model-based controller to be tracked. Preliminary experimental results are provided to establish the efficacy of this hierarchical control approach for trotting and walking gaits of a high-dimensional quadrupedal robot on unknown terrains and in the presence of disturbances.

Just Round: Quantized Observation Spaces Enable Memory Efficient Learning of Dynamic Locomotion

“…Compressed and embedded spaces: Previous work on learning in compressed or embedded spaces has mainly surrounded model-based techniques [22], [23], often learning an embedding directly from images [24], [25]. While modelbased learning has been shown to work well in some complex dynamic environments [26], model-free methods remain a popular choice in the dynamic locomotion community [16], [17].…”

“…While modelbased learning has been shown to work well in some complex dynamic environments [26], model-free methods remain a popular choice in the dynamic locomotion community [16], [17]. Others have turned to embedded and more descriptive action spaces [27], [28], [29], [30] and reduced order models [23] to enable more robust and sample efficient learning. However, these efforts have mainly ignored the impact of observation space compression on model-free learning.…”

Safety-Critical Control with Bounded Inputs via Reduced Order Models

“…This makes them challenging to control, and to attain formal guarantees of stable or safe evolution for the closed control loop. To tackle such complex control problems, simplified, reduced order models (ROMs) of the dynamic behavior are often utilized during controller synthesis with great practical success [1], [2]. Yet there is often a theoretic gap between behaviors certifiable on the ROM and the resulting behaviors observed on the full order system (FOS).…”

Data-driven Step-to-step Dynamics based Adaptive Control for Robust and Versatile Underactuated Bipedal Robotic Walking

“…Using a similar concept, a linear discrete reduced-order S2S dynamics model is learned in [19], [20] by extending a Hybrid-LIP based S2S approximation in [21]: both of which have a particular inputstate dynamics structure, where they consider the step size as the input to control horizontal center of mass (COM) states of bipedal walking. Additionally, the subspace approach in [22] has been used to obtain a data-driven reduced order model from the experimental data using the Hankel matrix [23], which ultimately realized quadrupedal locomotion experimentally. Although these methods use a robot-or problemspecific reduced-order model that can be used efficiently for online planning, they require offline data collection and computation to generate the data-driven approximation.…”