[Paper Review] A Data Driven Method of Optimizing Feedforward Compensator for Autonomous Vehicle

Tianyu Shi, Pin Wang+2 more|arXiv (Cornell University)|Jan 31, 2019

Vehicle Dynamics and Control Systems11 references4 citations

TL;DR

This paper proposes a data-driven feedforward compensator to enhance autonomous vehicle control robustness by modeling and compensating for system delays, parametric variations, and external disturbances. Using principal component analysis and a time-delay neural network to predict future errors, the method improves control accuracy and stability in simulation, demonstrating effectiveness in handling real-world uncertainties without relying on precise system models.

ABSTRACT

A reliable controller is critical and essential for the execution of safe and smooth maneuvers of an autonomous vehicle.The controller must be robust to external disturbances, such as road surface, weather, and wind conditions, and so on.It also needs to deal with the internal parametric variations of vehicle sub-systems, including power-train efficiency, measurement errors, time delay,so on.Moreover, as in most production vehicles, the low-control commands for the engine, brake, and steering systems are delivered through separate electronic control units.These aforementioned factors introduce opaque and ineffectiveness issues in controller performance.In this paper, we design a feed-forward compensate process via a data-driven method to model and further optimize the controller performance.We apply the principal component analysis to the extraction of most influential features.Subsequently,we adopt a time delay neural network and include the accuracy of the predicted error in a future time horizon.Utilizing the predicted error,we then design a feed-forward compensate process to improve the control performance.Finally,we demonstrate the effectiveness of the proposed feed-forward compensate process in simulation scenarios.

Motivation & Objective

• To address control inaccuracies in autonomous vehicles caused by time delays, parametric variations, and external disturbances such as road conditions and weather.
• To develop a model-free, data-driven approach that bypasses the need for precise system dynamics or disturbance models.
• To improve controller performance by predicting future control errors and applying feedforward compensation in real time.
• To validate the method in simulation under diverse dynamic and environmental conditions.

Proposed method

• Principal component analysis (PCA) is applied to extract dominant patterns from historical control data, identifying the most influential error sources.
• A time-delay neural network (TDNN) is trained to predict future control errors based on past input and state sequences.
• The predicted error is used as a feedforward compensation signal to adjust control commands before errors accumulate.
• The compensation process is integrated into the existing vehicle control architecture to enhance tracking performance.
• The method is trained and validated using simulation data reflecting real-world driving conditions and system variations.

Experimental results

Research questions

• RQ1 Can a data-driven approach effectively model and compensate for time delays and parametric variations in autonomous vehicle control?
• RQ2 How well can a time-delay neural network predict future control errors from historical data?
• RQ3 Does feedforward compensation based on predicted errors improve control accuracy and stability in simulation?
• RQ4 Can PCA effectively reduce the dimensionality of control error data while preserving critical dynamics?

Key findings

• The proposed data-driven feedforward compensator significantly reduces control error in simulation, improving tracking performance under various disturbances.
• The time-delay neural network achieved high accuracy in predicting future control errors, enabling effective pre-emptive compensation.
• PCA successfully identified the most influential error components, reducing data complexity without sacrificing predictive power.
• The method demonstrated robustness to external disturbances such as road surface changes and wind, maintaining stable control.