Technical Summary: C-ZUPT: Stationarity-Aided Aerial Hovering

Problem Statement

Autonomous systems operating in Global Navigation Satellite System (GNSS)-denied environments face a critical challenge: maintaining accurate state estimation over extended durations. While inertial navigation systems (INS) rely on high-frequency measurements, they are inherently prone to drift caused by sensor biases and noise. In the absence of external corrections (such as GNSS or visual updates), estimation errors accumulate unchecked, leading to filter divergence and loss of control.

Traditional solutions, such as Zero-Velocity Updates (ZUPT), are highly effective for surface-bound platforms (e.g., pedestrians, vehicles) by exploiting periods of physical rest. However, these methods are generally inapplicable to aerial platforms like quadrotors, which rarely achieve true zero-velocity even during hovering due to the need for continuous active control to counteract disturbances. Consequently, aerial systems lack a reliable, sensor-free mechanism to anchor their state estimation during stationary flight phases, leading to rapid accuracy degradation and increased energy consumption as controllers fight accumulating errors.

Methodology

The authors propose Controlled Zero-Velocity Update (C-ZUPT), a sensor-free state correction approach designed specifically for aerial navigation. The methodology integrates three core components:

1. Theoretical Framework and Modeling

The system models the quadrotor as a continuous-time dynamical system linearized around a hovering equilibrium point.

• Dynamics: A Newton-Euler framework describes the 6-DoF motion, accounting for thrust, gravity, and aerodynamic drag (neglected at low speeds).
• Control Architecture: The system employs a Linear-Quadratic-Gaussian (LQG) controller, combining a Linear Quadratic Regulator (LQR) for control with a Kalman Filter (KF) for state estimation.
• Battery Modeling: To evaluate energy efficiency, a Thevenin-based equivalent circuit model simulates battery dynamics, including voltage sag, state-of-charge (SoC) evolution, and transient responses to high-current demands.

2. The C-ZUPT Mechanism

The core innovation is a heuristic that identifies "quasi-static equilibria" during flight to trigger ZUPT corrections without requiring physical contact with a surface.

• Detection Logic: The algorithm monitors the $L_2$ norms of estimated body velocities ($\hat{v}_B$) and specific forces ($\tilde{f}$) over a sliding window ($K$ samples).
• Trigger Condition: When both norms fall below predefined thresholds ($\delta_{\hat{v}}$ and $\delta_{\tilde{f}}$), the system assumes a near-stationary state.
• Correction: Upon detection, the filter applies a pseudo-measurement update ($y_k = 0$) to the velocity states, effectively resetting the velocity error and constraining the growth of uncertainty. This is implemented as an "information aiding" mechanism, acting as a synthetic measurement derived from the physics of the hover equilibrium.

3. Signal Flow and Implementation

The approach accounts for real-world non-idealities in the control loop:

• Command Processing: LQR outputs are converted to rotor speeds via a non-negative least squares (NNLS) solver to ensure physical feasibility, followed by PWM generation.
• Actuator Dynamics: Rotor inertia and aerodynamic lags are modeled as first-order systems, creating a delay between command and actual thrust.
• Integration: The C-ZUPT update bypasses the standard prediction step in the KF when stationarity is detected, injecting the zero-velocity constraint directly into the estimation loop.

Key Contributions

1. Sensor-Free Aerial ZUPT: The paper introduces the first ZUPT implementation for aerial platforms that does not rely on surface contact, leveraging active control to maintain quasi-static equilibrium.
2. Drift-Resilient Navigation: A specific velocity correction method tailored for quadrotors that suppresses drift during stationary phases, validated through a publicly available framework.
3. Battery-Aware Estimation: An integrated circuit model that enables real-time energy monitoring, allowing the authors to quantify the relationship between estimation accuracy and power consumption.
4. High-Fidelity Motion Modeling: A comprehensive system model that couples LQG control with battery dynamics and actuator delays to simulate realistic flight conditions under sensor limitations.

Results and Analysis

Extensive simulations were conducted to evaluate the system under varying update rates ($\Gamma$) and disturbance levels, comparing the C-ZUPT aided system against an unaided baseline.

• Stability and Error Reduction: In scenarios with sparse GNSS updates (e.g., $\Gamma = 0.005$), the unaided system exhibited significant divergence in estimation and control errors. The C-ZUPT system maintained stability, reducing:
  • Tracking error by 28.8%
  • Attitude error by 21.6%
  • Control saturation time by 76.9%
  • Estimation uncertainty by 28%
• Control Effort: By correcting velocity drift, C-ZUPT reduced the control effort required to maintain the hover by approximately 31.9%.
• Energy Efficiency: The reduction in control effort translated directly to power savings. The C-ZUPT system demonstrated a 3% to 7% reduction in average power consumption and current draw compared to the baseline.
• Flight Time Extension: Due to improved efficiency, the estimated safe hover time (before reaching 30% battery SoC) increased by nearly 1 minute across all tested disturbance levels, representing a 3% to 6% improvement in overall system efficiency.

Significance and Claims

The authors claim that C-ZUPT provides a practical solution for extending the operational endurance of resource-constrained aerial systems in GNSS-denied environments. By leveraging the inherent stability of controlled hovering, the method mitigates filter divergence and enhances navigation stability without requiring additional sensors.

The paper posits that these improvements are critical for time-sensitive missions where energy efficiency and sustained flight are paramount, specifically citing applications in:

• Urban warfare and policing
• Rescue operations
• Persistent surveillance
• Infrastructure inspection
• Delivery services

The authors acknowledge limitations, noting that the method relies on mild dynamics (velocity errors must remain bounded) and requires careful tuning of heuristic parameters (thresholds and window sizes) to balance sensitivity against false positives. Nevertheless, the study concludes that C-ZUPT effectively transforms aerial stationarity intervals into a key asset for robust, sensor-free navigation.