DEVELOPMENT OF A MULTI-UNIT DEVICE FOR POSITIONAL TRACKING AND EXTERNAL LOAD ASSESSMENT IN FOOTBALL

Recently, there has been a growing interest in analyzing external load data in football, which includes metrics such as distance covered, speed, acceleration, and energy expenditure. Properly interpreting this data requires advanced knowledge and skills to extract useful insights. This creates a dual challenge: elite clubs may adopt the latest technologies but lack the expertise to fully analyze them, while non-professional and youth levels often face limitations in data collection. To address these challenges, this doctoral research presents the development and validation of wearable devices, consisting of a pair of smart shinguards equipped with a built-in GNSS (Global Navigation Satellite System) module designed to monitor and assess external load in football. The objective was to improve the accuracy of the starting version of the Device in the analysis of key athletic performance parameters such as distance covered, speed, acceleration, and energy expenditure. A set of four validation studies were conducted, comparing the device output to both gold-standard reference measures and industry-standard GPS vests. These tests were performed in controlled settings and ecological (soccer-specific) environments to ensure the device's performance across diverse dynamic scenarios. Key performance indicators evaluated included distance tracking, speed profiling, acceleration detection, and metabolic power estimation, which are critical for understanding player workload. Concerning distance covered, noise filtering, speed thresholding, and smoothing techniques were applied to the GNSS data in controlled settings, while track continuity was maintained through interpolation. In the ecological phase, similar steps were used, with additional processing for real-world signal disruptions and merging data from both shinguards. Data filtering algorithms were implemented to refine the interpretation of acceleration signals and energy consumption, enhancing the device’s ability to capture rapid directional changes and intense periods of play. For speed profiling, the initial speed calculations used positional differentiation, which were compared to a reference speed signal from a bike sensor in controlled settings. A new approach using Doppler shift was then applied, and the data was smoothed using a 5-step moving average to improve precision. In the ecological phase, real-world conditions were tested, and further filtering and synchronization of the speed signals from both Devices were conducted to refine the accuracy. The results showed that the Devices had strong alignment with the reference systems and GPS vests in terms of distance. The processing method provided accurate distance measurements in ecological environments, with results showing a MAPE (Mean Absolute Percentage Error) of 5.7\% and RMSE (Root Mean Square Error) of 264 m. About speed, the proposed method improved the global signal RMSE from 1.54 to 2.24 km/h in a sample-by-sample evaluation. In terms of detecting the number of accelerations and decelerations, the MAPE was reduced from 28\% to 17\% for accelerations, and from 51\% to 29\% for decelerations. Furthermore, the chosen processing methods for estimating energy expenditure resulted in a reduction in the related RMSE, decreasing from 12926 J/kg to 2622 J/kg. In conclusion, this research successfully improved the algorithms and validated a device consisting in a pair of smart shinguards with integrated GNSS for external load monitoring in football, enhancing the precision of key metrics such as distance, speed, accelerations/decelerations count, and energy expenditure across varied operative environments.