A Class of Axis-Angle Attitude Control Laws for Rotational Systems

1. Problem Statement

The paper addresses the challenge of designing robust attitude controllers for rotational systems (e.g., UAVs, satellites, microrobots) operating in 3D space. While traditional methods rely on quaternions or rotation matrices (geometric control), they possess inherent limitations:

• Quaternion Limitations: Standard quaternion-based control laws often use a proportional term scaled by $\sin(\Theta_e/2)$, where $\Theta_e$ is the rotation error. This causes the control effort to decrease as the error approaches $2\pi$ (the "unwinding" region), which is counter-intuitive for a feedback controller that should exert maximum effort when far from the target. Furthermore, to avoid the "unwinding" phenomenon (where the system takes a long path to stabilize), these controllers often require complex switching or hybrid schemes.
• Geometric Limitations: Geometric control laws implicitly choose the shortest rotational path to eliminate error. However, this shortest path is not always optimal for specific kinematic situations or performance objectives, particularly when angular velocity is high.
• Topological Obstruction: It is mathematically impossible to achieve global asymptotic stability on the Special Orthogonal group $SO(3)$ using continuous, time-invariant feedback. Existing solutions often involve discontinuous control laws or covering spaces, but these can introduce theoretical and experimental complexities.

2. Methodology

The authors propose a Generalized Axis–Angle Attitude Control Law that moves beyond quaternion-based formulations while retaining the benefits of the Euler axis–angle representation.

• Core Concept: The method defines the attitude error using a Scaled Euler Axis (SEA), denoted as $\boldsymbol{\alpha}_e = \gamma(\Theta_e)\mathbf{u}_e$.
  • $\mathbf{u}_e$: The unit Euler axis of rotation error.
  • $\Theta_e$: The rotation angle error (restricted to $[0, 2\pi)$).
  • $\gamma(\Theta_e)$: An extended class K function (a strictly increasing function where $\gamma(0)=0$).
• Control Law Formulation:
  The proposed torque input $\boldsymbol{\tau}_\gamma$ is given by:
  $$\boldsymbol{\tau}_\gamma = \mathbf{J}(k_\alpha \boldsymbol{\alpha}_e + k_\delta \dot{\boldsymbol{\alpha}}_e + k_\omega \boldsymbol{\omega}_e + \dot{\boldsymbol{\omega}}_d) + \boldsymbol{\omega} \times \mathbf{J}\boldsymbol{\omega}$$
  Where $\mathbf{J}$ is the inertia matrix, $\boldsymbol{\omega}_e$ is the angular velocity error, and $k_\alpha, k_\delta, k_\omega$ are positive gains.
• Stability Analysis:
  • The authors use Lyapunov stability theory combined with the properties of extended class K functions.
  • By restricting the domain of $\Theta_e$ to $[0, 2\pi)$ and defining the dynamics on an axis–angle covering (rather than directly on $SO(3)$), they ensure the existence of a unique equilibrium point ($\Theta_e = 0, \boldsymbol{\omega}_e = 0$).
  • Theorem 1 proves that if the gains satisfy specific conditions ($k_\delta > 0, k_\omega > 0, k_\alpha > \frac{k_\delta k_\omega}{4}$), the closed-loop system is globally asymptotically stable.
• Design Flexibility: The function $\gamma(\Theta_e)$ allows engineers to tune the control behavior. For instance, a sigmoid function can be used to provide linear behavior near equilibrium while saturating at large angles, preventing actuator saturation while maintaining high gain for large errors.

3. Key Contributions

1. Generalized Framework: Introduces a new class of control laws that generalize the Euler axis–angle representation, decoupling the control design from the rigid constraints of quaternion algebra.
2. Monotonic Control Effort: Unlike quaternion-based methods where control effort drops at large errors ($\Theta_e \approx 2\pi$), the proposed method ensures that the proportional control action increases as the system state moves further from the stable equilibrium. This is crucial for high-speed recovery maneuvers.
3. Stability Guarantees: Provides a rigorous mathematical proof for global asymptotic stability of a unique fixed point without relying on discontinuous switching for stability (though switching can still be used for performance optimization).
4. Compatibility with Switching: The framework is designed to work seamlessly with intelligent switching schemes (like Model Predictive Selection) that choose the optimal rotation direction based on angular velocity, avoiding the "unwinding" problem naturally.

4. Results

The authors validated the method through extensive numerical simulations and real-time outdoor flight experiments using a Crazyflie 2.1 quadrotor. The proposed controller ($\tilde{\boldsymbol{\tau}}_\gamma$) was compared against:

• A standard quaternion-based controller ($\boldsymbol{\tau}_b$).
• A high-performance geometric controller ($\boldsymbol{\tau}_g$).

Simulation Results (10,908 high-speed tumble-recovery maneuvers):

• Stabilization Time: For large initial rotation errors ($\Theta_{e,0} > 90^\circ$), the proposed method consistently achieved shorter stabilization times than both benchmarks.
• Control Effort: The proposed method required significantly less control effort (integral of squared torque) compared to the benchmarks, especially as the initial error increased.
• Behavior: The SEA norm ($\|\boldsymbol{\alpha}_e\|$) increased monotonically with the error angle, whereas the quaternion-based norm ($\|\mathbf{n}_e\|$) decreased as the error approached $2\pi$, confirming the theoretical advantage.

Experimental Results (30 real-world flight tests per controller):

• Performance: In real-world "tumble-recovery" tests (where the drone is thrown into a spin), the proposed controller achieved a mean stabilization time of 0.55 s, compared to 0.74 s (quaternion) and 0.61 s (geometric).
• Efficiency: The control effort required by the proposed method was approximately 25% lower than the geometric controller and significantly lower than the quaternion controller.
• Robustness: The method demonstrated superior performance under non-ideal conditions, including actuator saturation, model uncertainty, and external disturbances.

5. Significance

This work represents a significant advancement in rotational control theory and practice:

• Practical Utility: It solves the "unwinding" and "low-gain at large error" issues inherent in standard quaternion controllers, making it ideal for aggressive maneuvers like drone recovery from high-speed spins.
• Design Freedom: By allowing the selection of the function $\gamma(\Theta_e)$, it offers a flexible design space for tuning controllers to specific hardware constraints (e.g., actuator limits) and mission requirements.
• Validation: The combination of rigorous Lyapunov stability proofs with extensive real-world flight testing provides strong evidence that the theoretical benefits translate directly to improved physical performance.

In summary, the paper presents a mathematically sound and practically superior alternative to traditional quaternion and geometric attitude controllers, particularly for applications requiring rapid stabilization from large attitude errors.