Gait Feature Analysis and Recognition Under Coupled Weight Support and Bearing Conditions via a Domain-Adaptive Gradient Boosting Machine

Accurate recognition of gait characteristics under coupled body weight support and weight-bearing conditions is the core foundation for gravity adaptation strategy development.To solve this problem, this paper proposed a domain-adaptive gradient boosting machine (DAGBM) algorithm. The DAGBM algorithm added a cross-individual domain-adaptive regularization module to the standard GBM framework. It minimized both classification loss and maximum mean discrepancy (MMD) to align feature spaces across different subjects. It also worked with an individual–environment feature decoupling model to further improve generalization. Comparative experiments were conducted on a three-point suspension low-gravity test bed. A plantar pressure sensor with 48 points was used to collect gait data. This paper also extracted peak pressure, pressure–time integral, regional pressure ratio, and whole-foot center of pressure (CoP) trajectory standard deviation as features. DAGBM achieves 96.8% accuracy for seven body weight support levels, 97.4% accuracy for four weight-bearing levels, and 94.3% accuracy for all 28 fine-grained coupled conditions. DAGBM significantly outperforms both SGBoost and XGBoost algorithms, which exhibit favorable robustness under varied mechanical loading conditions and serve as a methodological reference for gait recognition research under aerospace-mimicking mechanical conditions.