Opening your car’s tailgate with a foot gesture works thanks to a dedicated capacitive sensor under the rear bumper. However, capacitive sensors tend to be unreliable due to the accumulation of dirt and debris.

In this project, we aimed to detect foot gestures using ultrasonic sensors already integrated into the bumper for systems such as Parking Distance Control (PDC) and Park Lane Assist (PLA).

To distinguish the correct gesture—intended to trigger the tailgate opening—from other movements of people or objects behind the vehicle, we employed a Convolutional Neural Network (CNN) architecture. The network compresses the time-series signals over a 3-second window and progressively aggregates features to produce a final descriptor.

We connected the ultrasonic sensors to an embedded system for real-time data processing and fed the signals into the trained neural network. The complete system was then installed on a Škoda Karoq, where we evaluated its performance and compared it with the original capacitive-sensor-based detector. A significant portion of our work focused on real-vehicle experimentation and fine-tuning the system’s structure and parameters. Our solution demonstrated significantly better performance compared to the capacitive sensor.

The partner of CTU was Škoda Auto.