Steering Wheel Hands-On Detection Using Multi-Sensor AI Signals

The patent combines multiple existing sensors (capacitive sensors, torque sensors, steering angle sensors) that are already present in the vehicle's steering system into a unified hands-on detection system. By merging the data from these sensors and processing them together through a machine learning model, the system achieves higher detection accuracy without adding significant hardware complexity, as the sensors are already part of the vehicle's standard equipment.

The patent introduces a machine learning model as an intermediary between the raw sensor data and the hands-on detection decision. This intermediary processes and integrates multiple sensor inputs (capacitive values, torque, steering angle) to produce a more accurate detection result. The machine learning model acts as a mediator that transforms complex multi-sensor data into reliable hands-on detection outcomes, improving measurement precision while keeping the overall system architecture manageable.

The patent applies machine learning models that have been pre-trained offline with extensive training data. The training process, which is computationally intensive, is performed beforehand rather than during real-time operation. During actual hands-on detection, the pre-trained model quickly processes sensor inputs, significantly reducing real-time computational complexity while maintaining high reliability. This preliminary action separates the heavy computational workload from the real-time detection process.

The system continuously processes sensor data and provides feedback to the machine learning model, which adjusts its predictions based on the incoming data patterns. The feedback mechanism allows the system to maintain high reliability by constantly evaluating new sensor readings against learned patterns, while the iterative nature of feedback processing is more computationally efficient than complete re-analysis of all data each time.

The patent transitions from static threshold-based rules to a dynamic machine learning approach that adapts to different driving situations and drivers. The machine learning model learns from training data encompassing various driving conditions, steering behaviors, and driver characteristics, enabling the system to dynamically adjust its detection criteria. This dynamic adaptation improves versatility while the model's automated learning process reduces the need for manual threshold adjustments, maintaining ease of operation.

The system changes the detection parameters from fixed threshold values to flexible, data-driven parameters learned by the machine learning model. Instead of using static thresholds for torque and capacitive values, the model learns optimal parameter ranges and relationships from training data, allowing it to adapt to different drivers and situations. This parameter transformation enables the system to handle diverse scenarios while the automated parameter learning maintains operational simplicity.