Graph-Based Self-Triggered Adaptive Fixed-Time Coordination for Autonomous Vehicle Formations With Safety Constraints

Resource-aware coordination of networked autonomous vehicles (AVs) requires fast convergence, safety preservation, uncertainty compensation, and efficient high-level command refreshes under limited onboard resources. This paper proposes a graph-based self-triggered adaptive fixed-time coordination scheme for longitudinal AV formations subject to prescribed spacing and velocity constraints. A smooth nonlinear transformation maps constrained physical variables into an unconstrained domain. To accommodate heterogeneous vehicle dynamics, the transformed uncertainty of each follower is represented over a common radial basis function (RBF) dictionary by a vehicle-dependent ideal coefficient vector. Neighboring followers exchange their current coefficient estimates over a communication graph; the resulting adaptive law produces both a Laplacian disagreement-dissipation term and a graph-weighted heterogeneity term in the Lyapunov analysis. A dynamic-surface controller with homogeneous fixed-time feedback is then combined with a self-triggered mechanism that refreshes and holds the transformed-domain command. The neural estimator and neural-weight exchange remain continuous, so the resource reduction claimed here concerns high-level transformed-command refreshes rather than neural communication. Lyapunov analysis establishes semi-global uniform ultimate boundedness, forward invariance of the original spacing and velocity constraints, practical fixed-time convergence with a heterogeneity-dependent residual bound, and exclusion of Zeno behavior.