Inertial measurement units (IMUs) for biomechanical analysis in sport: a review of applications, challenges and future directions

Purpose This paper aims to provide a critical, up-to-date synthesis of how inertial measurement units (IMUs) are used in sports biomechanics, clarifying their capabilities, common methodological pitfalls and the research gaps that must be closed for IMUs to become routine, high-value tools in performance analysis, injury prevention and rehabilitation. Design/methodology/approach A structured narrative review collated peer-reviewed studies (2010-April 2025) that: (i) describe IMU hardware or fusion algorithms; (ii) validate IMU-derived variables against gold-standard motion-capture or force-plate data; or (iii) apply IMUs to quantify technique, load or motor control in real-world sport. Key information was extracted on sensor specifications, calibration and processing pipelines, validation metrics and sport-specific outcomes, then organized into five thematic sections: fundamental principles, sport-by-sport applications, processing/validation methods, challenges and emerging directions (machine learning [ML], multimodal fusion and real-time feedback). Findings Across running, swimming, cycling, team, overhead, combat and niche sports, IMUs reliably capture spatiotemporal gait parameters (error < 3 %) and sagittal-plane joint angles (RMSE ≈ 3–8°). Their accuracy declines for nonsagittal kinematics and estimated kinetics, as demonstrated by validation studies reporting increased RMSEs (often >10°) and reduced intraclass correlation coefficients for joint angles in the frontal and transverse planes, largely due to soft-tissue artifact, sensor misalignment and magnetic interference. Recent machine-/deep-learning models and physics-informed hybrids now predict vertical ground-reaction forces with = 10 % normalized RMSE, and classification algorithms exceed 90 % accuracy for recognizing strokes, turns and strike types. Miniaturized, lower-power micro-electromechanical systems (MEMS) devices, integrated global positioning system – heart rate – electromyography (GPS–HR–EMG) wearables and real-time haptic/visual feedback prototypes are accelerating on-field adoption, yet standardized placement, calibration and reporting protocols remain scarce. Originality/value Unlike earlier reviews that focus on a single sport or device, this paper consolidates evidence across all major sporting domains, links specific sensor-fusion and modeling choices to measurable accuracy tradeoffs and distills actionable guidelines for researchers and practitioners. By mapping persistent obstacles – especially soft-tissue artifact and methodological heterogeneity – and pairing them with promising technological and analytical advances, this review outlines a clear research agenda to unlock the full potential of IMUs in sports science. Future directions include the adoption of ultra-low-power MEMS with enhanced bias stability, integration with multimodal systems and application of ML models capable of estimating vertical ground-reaction forces with normalized RMSEs below 10% and classifying sport-specific actions with over 90% accuracy. These innovations are expected to support fully autonomous, real-time biomechanical monitoring systems deployed across training and competition environments.