Proximally Optimal Predictive Control Algorithm for Path Tracking of Self-Driving Cars (2103.13240v2)

Abstract: This work presents proximally optimal predictive control algorithm, which is essentially a model-based lateral controller for steered autonomous vehicles that selects an optimal steering command within the neighborhood of previous steering angle based on the predicted vehicle location. The proposed algorithm was formulated with an aim of overcoming the limitations associated with the existing control laws for autonomous steering - namely PID, Pure-Pursuit and Stanley controllers. Particularly, our approach was aimed at bridging the gap between tracking efficiency and computational cost, thereby ensuring effective path tracking in real-time. The effectiveness of our approach was investigated through a series of dynamic simulation experiments pertaining to autonomous path tracking, employing an adaptive control law for longitudinal motion control of the vehicle. We measured the latency of the proposed algorithm in order to comment on its real-time factor and validated our approach by comparing it against the established control laws in terms of both crosstrack and heading errors recorded throughout the respective path tracking simulations.

• The paper introduces the POP controller that optimizes steering by minimizing prediction errors relative to lookahead points.
• It employs a proximity-based optimization loop that outperforms traditional controllers in lateral tracking accuracy.
• Simulation tests in CARLA confirm its real-time capability and reduced crosstrack and heading errors in autonomous driving.

Proximally Optimal Predictive Control for Autonomous Vehicle Path Tracking

The paper "Proximally Optimal Predictive Control Algorithm for Path Tracking of Self-Driving Cars" introduces a novel approach to the lateral control of autonomous vehicles using model-based predictive techniques. The authors focus on optimizing steering actions based on predicted vehicle locations, leading to improved path-tracking performance in self-driving scenarios. This work addresses the deficiencies found in traditional control approaches such as PID, Pure-Pursuit, and Stanley controllers, which often struggle with balancing tracking accuracy and computational efficiency.

Overview and Methodology

The core contribution of the paper is the development of the Proximally Optimal Predictive (POP) controller. This controller aims to enhance lateral path tracking by employing a proximity-based optimization loop that selects the optimal steering command within a neighborhood of the previous steering angle. This decision is made by minimizing the distance between the anticipated vehicle position and a lookahead point on the trajectory. The POP controller is designed to function efficiently in real-time, which the authors emphasize through simulation experiments using the CARLA Simulator.

The paper also details the contrast between the POP controller and traditional methods:

• PID Controller: Known for its versatility, the PID controller struggles with tracking efficiency when faced with dynamic changes in the system beyond its designed tuning range.
• Pure-Pursuit Controller: Although effective initially, this controller can generate erratic outputs towards trajectory ends due to an inability to handle missing lookahead points.
• Stanley Controller: While robust, the Stanley controller can manifest aggressive steering behaviors at low speeds and is sensitive to initial conditions.

Implementation and Results

The experimental validation conducted through a series of simulation tests underscores the practical efficacy of the POP controller. By maintaining a consistent vehicle and test environment, the paper compares the proposed approach against conventional controllers regarding crosstrack errors and heading errors. The POP controller exhibited superior tracking performance with reduced error metrics, significantly outperforming its counterparts.

The results also indicate that the POP approach achieves a real-time implementation capability, operating at a control loop rate over 300 Hz. This is a notable accomplishment given the algorithm's incorporation of prediction, optimization, and trajectory preprocessing within a Python-based framework.

Implications and Future Directions

From a practical standpoint, the implementation of the POP controller suggests several advancements in the field of autonomous vehicle control systems. The capacity to achieve precise real-time path tracking while maintaining computational efficiency makes this method attractive for embedded systems in autonomous vehicles. Moreover, the authors' algorithm facilitates a balance between the traditional simplicity of classical controllers and the computational demands of more advanced model predictive controls (MPC).

Theoretically, the methodology presents a compelling argument for integrating prediction-based processes directly into the control command's optimization. This framework could be extended in future research to encompass more complex vehicle dynamics models, possibly incorporating adaptive learning techniques to improve robustness across broader operational scenarios.

In conclusion, the paper effectively demonstrates that the Proximally Optimal Predictive Control algorithm is a substantial step forward in the domain of self-driving car control systems. Future work could explore hardware implementations, multi-agent scenarios, and the integration of this controller with other systems within autonomous platforms to enhance vehicle autonomy and safety on existing infrastructure. This contribution offers an exciting pathway for ongoing advancements in the field of autonomous vehicle control algorithms.