Flexible Gaits-Adaptive Pedestrian Dead Reckoning System: Precision Positioning Across Diverse Gaits

Abstract

With the rapid development of location-based services, the demand for accurate and robust indoor positioning has increased. Pedestrian dead reckoning (PDR) is the universal indoor positioning method, which provides self-reliant pedestrian positioning with relatively low deployment complexity. However, the growing demand for accurate and robust indoor positioning has highlighted the limitations of traditional PDR systems, which typically rely on a single default gait. A flexible gait-adaptive PDR system for multigait scenarios is proposed by considering eight common daily gaits and gait ambiguity. In this system, a gait-adaptive and multigait step length estimation (SLE) strategy is proposed in conjunction with a gait recognition (GR) routine. The GR routine with long short-term memory (LSTM) networks is taken as the most fundamental component of the system, determining the probabilities of pedestrian gait. Based on this, the contextual optimal SLE model is predicated on the highest probability gait by the gait-adaptive SLE (GA-SLE) strategy. An alternative method without the GR routine is provided by a multigait strategy, offering enhanced versatility in gait ambiguity environments. In addition, a waist-mounted experimental platform is designed and constructed to assess embeddability and positioning accuracy. Experimental results show that the multigait SLE strategy reduces average positioning errors by 46% to 12.80 m, while the GA-SLE strategy further improves accuracy, achieving a 79% reduction to 5.01 m.