FST.ai 2.0: An Explainable AI Ecosystem for Fair, Fast, and Inclusive Decision-Making in Olympic and Paralympic Taekwondo

TL;DR

FST.ai 2.0 introduces an explainable AI ecosystem for real-time, fair, and inclusive Taekwondo decision-making, integrating pose-based action recognition, uncertainty modeling, and visual explainability overlays to assist referees, athletes, and coaches. The system combines ST-GCN-based perception with transformer-oriented visual understanding, credal sets for epistemic uncertainty, and a governance layer to ensure transparency and accountability, while extending to Para-Taekwondo classification. Pilot deployments at international events demonstrate substantial reductions in decision review time (approximately 95%) and high referee trust (around 93%), with robust edge-enabled latency, and extensive dashboards for training analytics and policy insights. The framework envisions a scalable, federated, and ethically governed ecosystem that broadens impact to education, performance analytics, and federation-level decision support, emphasizing transparency, inclusivity, and human-in-the-loop control.

Abstract

Fair, transparent, and explainable decision-making remains a critical challenge in Olympic and Paralympic combat sports. This paper presents \emph{FST.ai 2.0}, an explainable AI ecosystem designed to support referees, coaches, and athletes in real time during Taekwondo competitions and training. The system integrates {pose-based action recognition} using graph convolutional networks (GCNs), {epistemic uncertainty modeling} through credal sets, and {explainability overlays} for visual decision support. A set of {interactive dashboards} enables human--AI collaboration in referee evaluation, athlete performance analysis, and Para-Taekwondo classification. Beyond automated scoring, FST.ai~2.0 incorporates modules for referee training, fairness monitoring, and policy-level analytics within the World Taekwondo ecosystem. Experimental validation on competition data demonstrates an {85\% reduction in decision review time} and {93\% referee trust} in AI-assisted decisions. The framework thus establishes a transparent and extensible pipeline for trustworthy, data-driven officiating and athlete assessment. By bridging real-time perception, explainable inference, and governance-aware design, FST.ai~2.0 represents a step toward equitable, accountable, and human-aligned AI in sports.

Figures (12)

• Figure 1: The interface displays key performance metrics for referees and athletes, including scoring latency, head-kick detection accuracy, referee response time, and decision consistency across matches. Users can filter by athlete, technique, or event type, enabling longitudinal performance analysis and targeted training interventions.
• Figure 3: Para Taekwondo Classification dashboard.
• Figure 4: Uncertainty-aware classtfication boundary betweenPara-Taekwonlo classes A6 and A7 based on Range of Motion (ROM) and Symmetry. The shaded red region represents high entropy, and the orange dot shows a borderline case flagged for expert.
• Figure 5: Modular system architecture -- showing arrows between modules, data layers, and feedback loops.
• Figure 6: Training feedback loop figure for this section as well -- depicting input clips → model overlays → referee response → feedback.