TL;DR
This paper introduces the HFO-LADRC, a novel controller designed to improve the lateral motion control of autonomous road sweepers. By proposing a new "heading-error-based model" and a corresponding first-order linear active disturbance rejection controller, the HFO-LADRC effectively addresses coupled lateral and heading errors, demonstrating superior robustness and stability against various uncertainties compared to traditional methods.
What Problem Does This Paper Address?
Achieving robust and stable motion control is a critical challenge for autonomous systems, especially in the face of inherent uncertainties like sensor noise, parameter variations, and external disturbances. For autonomous road sweepers, precise lateral control is essential for efficient operation and safety, ensuring the vehicle accurately follows its intended path. Existing control methods often struggle to simultaneously manage both lateral position error and heading error, particularly when dealing with real-world complexities. This paper tackles this challenge by developing a controller specifically designed to reject disturbances and maintain stability for autonomous road sweepers, even when faced with uncertainties in parameters like wheelbase and steering ratio.
Methods
The core of this work lies in the development of a Heading-Error-Based Model for lateral motion. This new model aims to eliminate both lateral error ($y_e$) and heading error ($\phi_e$) simultaneously. To achieve this, the authors introduce a new state variable, $z$, which combines these errors using a cotangent hyperbolic function. The stability and convergence of this model are mathematically proven using a Lyapunov function.
Based on this first-order heading-error-based model, a First-Order Linear Active Disturbance Rejection Controller (HFO-LADRC) is designed. This controller comprises three key modules:
- Linear Tracking Differentiator (LTD): Smooths the input signal and generates derivatives.
- Linear Extended State Observer (LESO): Estimates the system's internal dynamics and external disturbances as a total disturbance.
- Linear State Error Feedback (LSEF): Uses the estimated states and disturbance to generate the control signal.
The controller is designed to be linear and first-order, aligning with the proposed model. For simulation accuracy, the fourth-order Runge-Kutta method is employed as the differential system solver.
Datasets
The paper does not use pre-existing datasets for training. Instead, it relies on simulated scenarios designed to mimic real-world conditions. These scenarios include:
- A typical ring scenario with straight lanes and semicircles.
- A double-lane-changing obstacle avoidance scenario.
These scenarios were designed referencing the ISO3888-2:2011 standard. The simulations were conducted in a Matlab/Simulink environment.
Key Insights and Ideas
- Unified Error Handling: The novel "heading-error-based model" effectively couples lateral and heading errors, allowing a single controller to address both simultaneously, which is crucial for precise path following.
- ADRC for Robustness: Leveraging the Active Disturbance Rejection Control (ADRC) framework, particularly its linear variant, proves highly effective in rejecting external disturbances and parameter uncertainties, a common issue in real-world autonomous systems.
- First-Order Simplicity: Designing a first-order LADRC based on the first-order heading-error model simplifies the controller architecture while maintaining performance, making it potentially more computationally efficient.
- Lyapunov Stability Proof: The rigorous mathematical proof using a Lyapunov function provides strong theoretical backing for the convergence of both lateral and heading errors, increasing confidence in the controller's stability.
- Uncertainty as a Key Focus: The paper explicitly tests the controller's performance under various uncertainties (wheelbase, steering ratio, Gaussian white noise), highlighting its practical relevance for autonomous vehicle applications.
Key Results
The HFO-LADRC demonstrated superior performance compared to a typical pure pursuit controller and a classical second-order LADRC (SO-LADRC) across various test scenarios and uncertainty conditions.
- Wheelbase Uncertainty: In the ring scenario, the HFO-LADRC achieved a peak lateral error of 0.0342 m, significantly lower than the SO-LADRC (1.1 m) and pure pursuit (0.457 m). Similar improvements were observed in the double-lane-changing scenario.
- Steering Ratio Uncertainty: The HFO-LADRC consistently showed the lowest peak lateral error (e.g., 0.0462 m in the ring scenario) compared to SO-LADRC (1.2 m) and pure pursuit (0.556 m).
- Gaussian White Noise Disturbance: Even with added noise, the HFO-LADRC maintained strong performance, with a peak lateral error of 0.031 m in the ring scenario, outperforming SO-LADRC (1.058 m) and pure pursuit (0.418 m).
- Robustness and Stability: Across all tested uncertainties and scenarios, the HFO-LADRC exhibited significantly stronger robustness and stability, with lateral error reduction capacities often an order of magnitude better than the compared methods.
Limitations and Future Work
- Limitations: The study primarily relies on simulations. While the scenarios are designed to be realistic, real-world testing with actual road sweepers is necessary for full validation. The paper focuses on specific uncertainties; other factors like sensor degradation or actuator limitations might require further investigation.
- Future Work: The authors suggest extending the research to more complex test scenarios and performing real-vehicle tests. They also propose investigating a wider range of uncertainties to further enhance the algorithm's robustness. Additionally, exploring the dynamic characteristics of actuators for emergency situations (e.g., auto emergency braking) and studying multi-system joint operations (e.g., integrating motion control with environment perception and decision-making) are identified as promising future directions.
Summary updated