Fuzzy Logic based Autonomous Parking Systems - Part IV: A Multiple-Model Adaptive Neural-Fuzzy Controller

Yu Wang, Xiaoxi ZhuarXiv 2019 · OpenReview

TL;DR

This paper introduces a novel Multiple-Model Adaptive Fuzzy Logic Controller (MM-AFLC) for autonomous parking systems. By combining neural network-based non-linear system identification with a multi-model adaptive fuzzy logic approach, the controller achieves robust and efficient parking performance, even under challenging conditions like large ground friction.

What Problem Does This Paper Address?

Autonomous driving, particularly parking, is a complex task that has seen significant research. While PID controllers are common in industry due to their balance of performance and cost, they struggle with non-linear systems and unpredictable disturbances. Fuzzy logic controllers offer a more robust alternative, simulating human reasoning. However, traditional fuzzy controllers can suffer from oscillations and slow convergence, especially with varying plant models and external factors. This paper aims to bridge this gap by developing a controller that is robust, adaptive, and achieves faster, smoother convergence for autonomous parking, overcoming limitations of previous approaches.

Methods

The proposed solution employs a two-step approach:

  1. Off-line Neural Network Identification: A neural network is used to identify the non-linear dynamics of the vehicle's plant model. This is crucial because deriving an exact mathematical model for vehicles is difficult due to their inherent non-linearity and complexity. The neural network acts as a universal approximator, learning the system's behavior without prior knowledge.
  2. Multiple-Model Adaptive Fuzzy Logic Control (MM-AFLC): This is the core of the proposed controller. It builds upon adaptive control principles by:
  3. Adaptive Control with Model Identification: It uses a reference model to define the desired system behavior and an adaptive law (Lyapunov-based) to adjust controller parameters to minimize the error between the actual and desired states. The neural network identification is integrated here for improved stability.
  4. Multiple Models Adaptation: Instead of relying on a single fuzzy controller, the MM-AFLC uses multiple fuzzy identification models. These models are combined using a convex hull approach, effectively restricting the search space for optimal parameters and accelerating convergence. This multi-model strategy enhances robustness and reduces oscillation.

The controller is designed to handle both vehicle speed and steering wheel control.

Datasets

Not discussed in the paper. The experiments are described as simulations using a "plant model" and comparative tests against other controllers.

Key Insights and Ideas

  • Synergy of NN and Fuzzy Logic: Combining neural networks for system identification with fuzzy logic for control offers a powerful approach to handle complex, non-linear systems where precise models are hard to obtain.
  • Multiple Models for Faster Convergence: The key innovation is the "multiple models" aspect. By considering a set of fuzzy identification models within a convex hull, the controller can converge to the optimal solution much faster and with less oscillation compared to single-model adaptive approaches.
  • Robustness through Adaptation: The adaptive nature of the controller, coupled with the multi-model strategy, allows it to maintain performance even under significant external disturbances like large ground friction, which is a critical factor in real-world autonomous driving.
  • Generalizability: The approach is designed to be generalized, meaning the core controller can be applied to different systems with minimal modifications after the initial neural network identification step.

Key Results

The paper presents simulation results comparing the proposed MM-AFLC with three other fuzzy controller-based approaches (FBOS, HFC, and HFC-FTD).

  • Improved Parking Performance: Under conditions of large ground friction, the MM-AFLC and HFC-FTD were able to complete the parking process, while FBOS and HFC struggled.
  • Faster and Smoother Parking: The MM-AFLC demonstrated a significantly faster and smoother parking trajectory compared to HFC-FTD, as evidenced by the more scattered dots in the simulation path.
  • Superior Angular Speed Tracking: Figure 4 shows that the MM-AFLC exhibits a much smaller tracking error for angular speed, indicating greater robustness and superior performance under challenging friction conditions.

Limitations and Future Work

  • Bounded Error: The authors acknowledge a bounded error between the actual fuzzy control signal and the optimal value. While this error is bounded, it can still impact overall system performance.
  • Future Work:
    • Minimizing Bounded Error: Future research could focus on minimizing this bounded error by adjusting the adaptive laws.
  • Real-World Industrial Application: Further work is needed to address practical challenges for real industrial applications, such as implementation cost and operational efficiency.

Summary updated

Abstract

In this paper, a Multiple Models Adaptive Fuzzy Logic Controller (MM-AFLC) with Neural Network Identification is designed to control the unmanned vehicle in Intelligent Autonomous Parking System. The objective is to achieve robust control while maintaining a low implementation cost. The proposed controller design incorporates the following control theorems -- non-linear system identification using neural network, fuzzy logic control, adaptive control as well as multiple models adaptation. Such integration ensures superior performance compared to previous work. The generalized controller can be applied to different systems without prior knowledge of the actual plant model. In the intelligent autonomous parking system, the proposed controller can be used for both vehicle speed control and steering wheel turning. With a multiple model adaptive fuzzy logic controller, robustness can be also assured under various operating environments regardless of unpredictable disturbances. Last but not least, comparative experiments have also demonstrated that systems equipped with the new controller are able to achieve faster and smoother convergence.
BibTeX
@misc{wang2019fuzzy,
  author = {Yu Wang and Xiaoxi Zhu},
  title = {{Fuzzy Logic based Autonomous Parking Systems - Part IV: A Multiple-Model Adaptive Neural-Fuzzy Controller}},
  year = {2019},
  eprint = {1911.02703},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1911.02703},
}
Figure