Quaternion-Based Hybrid Control for Robust Global Attitude Tracking

Christopher G. Mayhew, Ricardo G. Sanfelice, Andrew R. Teel

IEEE Transactions on Automatic Control · 2011 · 421 citations · 37 references

Concepts

TL;DR

Attitude control of rigid bodies faces topological constraints that limit global stabilization. The paper proposes a quaternion‑based hybrid feedback scheme to achieve global attitude tracking under full state, attitude‑only, and biased angular‑velocity measurement scenarios. The scheme employs a dynamic hybrid controller with hysteresis‑based switching and, when needed, an observer that incorporates a quaternion filter and bias estimator. Simulations confirm that the hysteresis mechanism eliminates unwinding and reduces sensitivity to small measurement noise, validating the hybrid controller’s effectiveness.

Abstract

It is well known that controlling the attitude of a rigid body is subject to topological constraints. We illustrate, with examples, the problems that arise when using continuous and (memoryless) discontinuous quaternion-based state-feedback control laws for global attitude stabilization. We propose a quaternion-based hybrid feedback scheme that solves the global attitude tracking problem in three scenarios: full state measurements, only measurements of attitude, and measurements of attitude with angular velocity measurements corrupted by a constant bias. In each case, the hybrid feedback is dynamic and incorporates hysteresis-based switching using a single binary logic variable for each quaternion error state. When only attitude measurements are available or the angular rate is corrupted by a constant bias, the proposed controller is observer-based and incorporates an additional quaternion filter and bias observer. The hysteresis mechanism enables the proposed scheme to simultaneously avoid the "unwinding phenomenon" and sensitivity to arbitrarily small measurement noise that is present in discontinuous feedbacks. These properties are shown using a general framework for hybrid systems, and the results are demonstrated by simulation.

References

37