Xuheng Wang

The paper introduces a novel adversarial attack that uses static, view-dependent camouflage on a vehicle to induce consistent feature drift, causing autonomous systems to predict false, yet plausible, trajectories like unnecessary cut-ins.