Real-Time LPV-Based Non-Linear Model Predictive Control for Robust Trajectory Tracking in Autonomous Vehicles

Nitish Kumar, Rajalakshmi Pachamuthu · 2025

Read Real-Time LPV-Based Non-Linear Model Predictive Control for Robust Trajectory Tracking in Autonomous Vehicles by Nitish Kumar, Rajalakshmi Pachamuthu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-04684.txt.

Introduction

Nitish Kumar and Rajalakshmi Pachamuthu build a control system that keeps a driverless car glued to its planned path, in a 2025 paper. Their method looks ahead, predicts the car's motion, and picks the best steering and speed moves many times a second. The car's dynamics are written in a form whose terms shift with speed and curve, and a tuning rule adjusts the controller's weights as the road bends.

The system runs on a robot software platform that splits sensing and control work to keep delays low. In simulations and real runs on set routes, the car held its line with tiny errors, even in sharp maneuvers and at high speed. Simulated and real behavior matched closely. The authors plan dynamic weight tuning and links to cooperative driving, where cars share plans.

Catalog record

CREATORNitish Kumar, Rajalakshmi Pachamuthu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-04684

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Real-Time LPV-Based Non-Linear Model Predictive Control for Robust Trajectory Tracking in Autonomous Vehicles Nitish Kumar, Rajalakshmi Pachamuthu Abstract This paper presents the development and implementation of a Model Predictive Control (MPC) framework for trajectory tracking in autonomous vehicles under diverse driving conditions. The proposed approach incorporates a modular architecture that integrates state estimation, vehicle dynamics modeling, and optimization to ensure real-time performance. The state-space equations are formulated in a Linear Parameter Varying (LPV) form, and a curvature-based tuning method is introduced to optimize weight matrices for varying trajectories. The MPC framework is implemented using the Robot Operating System (ROS) for parallel execution of state estimation and control optimization, ensuring scalability and minimal latency. Extensive simulations and real-time experiments were conducted on multiple predefined trajectories, demonstrating high accuracy with minimal cross-track and orientation errors, even under aggressive maneuvers and high-speed conditions. The results highlight the robustness and adaptability of the proposed system, achieving seamless alignment between simulated and real-world performance. This work lays the foundation for dynamic weight tuning and integration into cooperative autonomous navigation systems, paving the way for enhanced safety and efficiency in autonomous driving applications. I Introduction Autonomous vehicles (AVs) have revolutionized the transportation industry by offering safer, more efficient, and intelligent mobility solutions. A crucial aspect of autonomous driving is path tracking, which ensures that the vehicle follows a desired trajectory accurately and smoothly under varying driving conditions. Effective path tracking not only requires precise control of vehicle dynamics but also demands real-time computational efficiency to adapt to the constantly changing environment. In this regard, Model Predictive Control (MPC) [ 1 ] [ 2 ] has emerged as a prominent approach due to its predictive capabilities, optimization-based formulation, and ability to handle constraints on system states and control inputs. While MPC has been successfully applied for path tracking, [ 3 ] [ 6 ] [ 4 ] existing solutions often fall short of achieving real-time performance when integrated with multiple sensors and complex vehicle dynamics. Traditional approaches typically assume fixed speed control with minimal adaptability, which does not reflect the dynamic behavior observed in manual driving [ 5 ] [ 7 ] . In manual mode, human drivers intuitively adjust the vehicle’s speed based on environmental factors, road conditions, and personal comfort preferences. Emulating this behavior in autonomous systems introduces significant computational challenges, particularly when real-time sensor integration, [ 8 ] next-state prediction, and optimization need to be performed simultaneously. To address these challenges, this paper presents a novel MPC-based path tracking framework capable of operating in two distinct speed modes—fixed-speed mode and variable-speed mode—that replicate the flexibility of manual driving. In the proposed framework: • The fixed-speed mode maintains a constant speed along the desired trajectory, ensuring stability and predictability under uniform road conditions. • The variable-speed mode dynamically adjusts the vehicle speed in response to external factors such as sharp turns, path curvature, and real-time sensor inputs, mimicking human-like driving behavior. To achieve real-time computational efficiency, a key challenge in implementing MPC, this work leverages parallelized computations to optimize the controller’s performance. By dividing tasks such as next-state prediction, cost function evaluation, and optimization into parallel processes, the proposed framework effectively reduces computation time without compromising accuracy. The entire system is implemented within the Robot Operating System (ROS) environment [ 4 ] , which facilitates seamless communication between the controller, sensors, and actuators. Specifically, the system integrates data from multiple sensors in real-time, processes the inputs to generate control parameters (acceleration and steering), and actuates the vehicle with minimal latency. The key contributions of this paper are as follows: • Flexible Speed Control Modes: A method to implement both fixed and variable-speed modes in autonomous driving, enabling the vehicle to adjust its speed dynamically, similar to manual driving behavior. • Parallelized MPC Computations: A computationally efficient MPC framework that parallelizes next-state prediction, cost function evaluation, and optimization to enhance real-time performance. • Real-Time Sensor Integration: Real-time fusion of sensor data and integration with ROS to provide accurate control inputs for steering and acceleration, ensuring seamless actuation. • Reduced Computational Load: The proposed framework significantly reduces the computational load by optimizing resource utilization, enabling smooth and responsive vehicle control in dynamic environments. The proposed system is validated through real-time experiments, demonstrating its effectiveness in achieving precise path tracking under varying speed modes. The results highlight the system’s ability to adapt to sharp turns, curvatures, and dynamic conditions while maintaining low computational latency and high accuracy. II Related Works Nonlinear Model Predictive Control (NMPC) has been extensively explored in recent years for autonomous navigation, path tracking, and collision avoidance due to its ability to handle dynamic constraints and uncertainties. Various studies have demonstrated its versatility in solving complex driving tasks. In [ 9 ] , Anil et al. compared MPC with conventional PID controllers for trajectory tracking, showcasing that MPC achieved superior path-following performance and stability in autonomous driving systems. Similarly, Li et al. , in [ 10 ] , proposed an MPC-based navigation framework that incorporates sensor fusion, neural networks, and path exploration, enabling robust performance in uncertain environments. MPC has also been applied for longitudinal control, addressing smooth speed adjustments while accounting for environmental constraints. In [ 11 ] , Mekala et al. used MPC with LiDAR inputs to control vehicle acceleration and deceleration, ensuring a stable response under varying speeds and obstacles. Similarly, Kim et al. , in [ 12 ] , integrated RL with MPC to optimize real-time highway navigation. Their hybrid approach successfully reduced computational time by over 90% while maintaining trajectory accuracy, demonstrating the effectiveness of machine learning techniques in augmenting traditional MPC. Collision avoidance remains a crucial challenge in autonomous systems, and several studies have proposed solutions by integrating MPC with advanced techniques. Yang et al. , in [ 13 ] , combined an adaptive Artificial Potential Field (APF) method with MPC to generate smooth, collision-free trajectories. By using triangular collision constraints, the computational burden was reduced while maintaining feasibility under dynamic environments. Another notable work, Distributed Cooperative MPC [ 14 ] by Mohseni et al. , proposed a distributed control approach for multi-vehicle systems. Their framework ensured real-time cooperative lane-switching and intersection maneuvers, highlighting MPC’s scalability and efficiency in handling multi-agent control. To address proactive control, Yoon et al. employed Gaussian Process Regression (GPR) in [ 15 ] . By predicting cut-in vehicle behaviors, their method improved prediction accuracy and enabled the autonomous vehicle to optimize its control strategy, reducing sudden maneuvers. This predictive capability was further extended in [ 16 ] by Li et al. , where computational efficiency was enhanced using dimension reduction strategies, facilitating real-time deployment in dynamic urban environments. MPC has also been applied for multi-vehicle cooperative systems. Li et al. , in [ 17 ] , designed a framework that incorporated driver behavior characteristics to ensure safer and more human-like cooperative driving. Similarly, the cooperative approach was expanded in [ 14 ] , where distributed strategies enabled vehicles to navigate efficiently in complex traffic situations. Beyond ground vehicles, MPC applications extend to maritime navigation. Tsolakis et al. , in [ 18 ] , proposed a trajectory optimization method compliant with International Maritime Traffic Rules (COLREGs). Their approach combined rule-based constraints with MPC to ensure

Read the complete text