Time delay estimation-based fixed-time controller for autonomous vehicles

A key challenge in autonomous vehicle lateral control is the trade-off between the computational simplicity of geometric controllers and the high-speed accuracy of dynamic-model-based approaches. This paper addresses this challenge through a time delay estimation -based fixed-time control architecture built upon a new geometric framework. First, the predictive and path-smoothing characteristics of Pure Pursuit are combined with the cross-track-error correction of the Stanley method to form a unified geometric controller. The resulting error dynamics are designed such that the lateral tracking error converges faster than the heading error, thereby mitigating the corner-cutting tendency of conventional geometric controllers. Second, time delay estimation is introduced as a real-time compensation mechanism to estimate and reject the lumped effects of unmodelled vehicle dynamics, parameter uncertainties, road conditions, and external disturbances. Third, fixed-time stability is established, guaranteeing convergence within a bounded time independent of the initial conditions. Furthermore, the proposed geometric formulation is transformed into an LPV kinetic representation, establishing a direct relationship between controller parameters and the vehicle dynamic model and enabling systematic stability-oriented tuning across different speeds and vehicle types. Simulations, including demanding double-lane-change manoeuvres above 100 km/h, demonstrate improved tracking accuracy, robustness, and stability compared with LPV-MPC, gain-scheduled H∞, and nonlinear controllers.