Electrokinetic self-cleaning of solid–fluid contaminants on perception sensors in autonomous vehicles

This study presents an electrokinetically actuated self-cleaning methodology to ensure the robust operation of vision-based object recognition systems in autonomous vehicles. Surface contamination of optical perception sensors substantially compromises image fidelity and detection reliability. In real-world operational environments, heterogeneous mixtures of solid particulates and liquid-phase contaminants (e.g., moisture and oil films) frequently co-occur, necessitating an integrated removal strategy to maintain sensor accuracy and stability. To overcome these limitations, we propose an electric-field-driven cleaning approach employing a multifunctional electrode architecture that synergistically combines electrowetting-on-dielectric (EWOD) and dielectrophoresis (DEP) within a single actuation framework. First, the electrohydrodynamic and dielectrophoretic responses of droplets and particulates were characterized under time-varying electric fields. Subsequently, mixed-phase contaminants, including aqueous droplets, oil residues, and fine dust particles, were systematically applied to the sensor surface. Cleaning efficacy was quantitatively assessed by measuring the areal reduction of contamination pre- and post-actuation using image-based analysis techniques. Finally, the optimal actuation parameters were applied in a simulated object recognition task, demonstrating a significant enhancement in vision-based detection accuracy.