Wearable sensor-guided hierarchical reinforcement learning for adaptive motor skill correction and injury-risk assessment in physical education

Physical education is essential for developing motor skills, improving physical fitness, and promoting long-term health. However, conventional training methods primarily depend on manual instructor observations, making it difficult to detect incorrect movement patterns, thereby increasing the likelihood of sports-related injuries. To overcome these limitations, this research proposes a Wearable Sensor-Guided Fossa Optimizer tuned-Option-Critic Hierarchical Reinforcement Learning (FO-OC-HRL) model for adaptive motor skill correction and injury-risk assessment in physical education. The proposed model acquires Wearable Motor Skill Assessment data containing 7000 student records, including accelerometer, gyroscope, heart rate, and electromyography (EMG) signals, to continuously monitor learners’ movement dynamics and physiological conditions during training sessions. Initially, raw sensor signals are denoised using a Butterworth filter and standardized through Z-score normalization. Subsequently, Discrete Wavelet Transform (DWT) is employed to extract discriminative temporal and frequency-domain features representing movement quality, muscle activation, and biomechanical characteristics. The extracted features are then processed by the FO-OC-HRL model, where the FO automatically tunes the hyperparameters of an OC-HRL architecture, while OC-HRL performs hierarchical policy learning to achieve efficient policy learning and faster convergence. The high-level policy autonomously selects appropriate motor skill correction strategies, while low-level option policies generate personalized corrective actions and adaptive exercise recommendations based on the learner’s performance. The FO-OC-HRL model achieved 98.27% Accuracy, demonstrating superior predictive performance and computational efficiency, which was evaluated using Python 3.11. FO-OC-HRL provides personalized decision support using wearable sensing and adaptive HRL to improve motor skills, reduce injury risks, and enhance physical education.