Tire Blowout Detection Algorithm for Vehicles Based on PSO-SVM and Multi-Sensor Signals Determination

This study proposes a robust tire blowout identification algorithm for autonomous vehicles by integrating multi-sensor fusion with a Particle Swarm Optimized Support Vector Machine (PSO-SVM). Through a comparative analysis of vehicle dynamic responses under blowout versus non-blowout conditions within the dataset, 22 signals exhibiting significant differential trends were initially considered. The ReliefF algorithm selected five discriminative features for training the SVM. Signal preprocessing, such as moving average filtering and steering-induced yaw rate compensation, was applied to enhance feature stability and reliability. A comprehensive evaluation metric incorporating false negatives, false positives, and detection latency was designed as the optimization target. The PSO algorithm optimized SVM hyperparameters (penalty factor c and kernel parameter g) to enhance classification performance. This optimization strategy effectively balances detection accuracy and timeliness under complex driving scenarios. Validation across 90 co-simulated scenarios—covering blowouts, potholes, and bumps —achieved 0 false alarms or missed detections. The proposed framework offers a reliable perception foundation for stability control in blowout conditions.