Packet-Loss Resilient Heavy-Duty Truck State Estimation via Event-Triggered Data-Driven Filtering in V2X Networks

Intelligent heavy-duty trucks require precise preceding vehicle state estimation for safety and efficiency, but face challenges from inherent structural complexities, model uncertainties, and unreliable V2X communication. Traditional model-based filters struggle with computational complexity and packet loss, while periodic V2X transmissions waste bandwidth. This paper proposes a Strong Tracking Event-Triggered Data-Driven Consensus Filter (ST-ETDDCF) for resilient state estimation under packet loss. Initially, the nonlinear truck system is converted into a pseudo-linear data model through dynamic linearization. An adaptive event-triggered mechanism then reduces communication load. Data is transmitted only when prediction errors exceed the threshold. The Strong Track Filter enhances robustness via time-varying fading factors. When the innovation is biased, the covariance of the prediction will be increased. Crucially, the framework compensates for packet loss using historical measurements and derives an estimation error upper bound. Simulations validate that ST-ETDDCF outperforms others in lateral velocity and yaw rate tracking accuracy under identical conditions, demonstrating superior stability and resilience for systems.