Resource-Efficient Path-Following Control for a Self-Driving Car in a Networked Control System

Abstract

In recent years, in-vehicle networks are increasingly being incorporated to self-driving cars in order to interconnect spatially distributed devices such as sensors, actuators, and controllers, leading to networked control systems (NCS). The main aim of this work is to reduce the use of resources in a NCS (bandwidth, device batteries) while maintaining an accurate path following for a self-driving car. Some typical network-induced drawbacks such as time-varying delays, packet dropouts and packet disorder will also be coped with. In order to reach the goals, a systematic integration of periodic event-triggered sampling techniques, packet-based control strategies, and state estimation methods is proposed. A novel non-uniform dual-rate extended Kalman filter (NUDREKF) is formulated to estimate the system state at fast, control rate from scarce slow-rate measurements. Due to its mathematical simplicity and low computational cost, the dynamic control law is designed from an inverse kinematic bicycle model and a proportional feedforward controller. Interestingly, optimal parameters for the event-triggered conditions are reached, leading to a satisfactory trade-off between resource savings and control performance. Simulation results for a real trajectory considering actual limitations for the actuators reveal the benefits of the control proposal compared to a conventional control approach.