Electronic travel aids are pivotal for the independent mobility of the visually impaired. While Vision-Language Models (VLMs) offer rich environmental understanding, they often suffer from excessive false positives in dynamic scenarios, leading to cognitive overload. To address this, we present ForeSightGuide, an anticipatory assistive guidance framework that couples semantic scene understanding with predictive hazard assessment. Unlike reactive systems, ForeSightGuide leverages the reasoning capabilities of VLMs to anticipate obstacle motion, effectively filtering out non-threatening objects to provide concise, actionable guidance. To validate our approach, we introduce a novel dataset captured in complex, dynamic real-world traffic scenes, designed to benchmark predictive capabilities. Extensive experiments on both public benchmarks and our proposed dataset demonstrate that ForeSightGuide achieves state-of-the-art performance. Notably, it significantly mitigates information overload by reducing redundant alerts to 0.299 per guidance output while maintaining a low missed-hazard rate of 0.112, proving its efficacy for safe walking assistance.
Zhiyuan Wang, Xu Li, Shikang Guo et al.· 0 citations
Assist-as-needed (AAN) assistance can encourage active participation during human walking. However, individuals exhibit diverse walking patterns, which makes it challenging for exoskeleton to provide personalized assistance. This paper presents an innovative adaptive AAN control strategy with hybrid torque fusion, including two modules of hybrid torque fusion estimation and adaptive control to promote voluntary participation. Specifically, the hybrid torque fusion module fuses torque estimated from surface electromyography (sEMG) signals with dynamic model estimation using regularized particle filter fusion (RPFF) algorithm to evaluate the human’s active participation. The control module incorporates a globally continuous extended assistance level function (EALF) that integrates joint motion tracking error, human-robot interaction force, and voluntary deficit to quantify subject performance and enable smooth transitions between four training modes, thereby continuously adjusts the torque output. Comprehensive comparisons with existing AAN strategies under multiple scenarios demonstrate that the proposed method improves trajectory tracking accuracy by approximately 16.1% and reduces human-robot interaction force by 4.01% during slope walking and simulated gait impairment. NASA-TLX and Likert scale assessments further validate the method’s effectiveness in enhancing active participation and wearing comfort. Note to Practitioners—To address insufficient active participation and unsmooth assistance in walking rehabilitation using ankle exoskeleton, caused by incomplete quantification of human motor ability, this study introduces a human active torque fusion mechanism. We developed a personalized human-robot collaboration model and proposed an AAN control algorithm based on an assistance function. This algorithm can potentially be transferred or extended to other rehabilitation robotic platforms. Although the system demonstrates promising application potential, this study acknowledges certain limitations, such as a small sample size and insufficient exploration of diverse patient population needs. Future research can build on this work by integrating additional sensing technologies, optimizing control strategies to enhance system adaptability, and exploring more efficient adaptive or online learning methods to simplify the parameter calibration process before training. This study offers the potential to significantly improve daily living for individuals with mobility limitations, laying the groundwork for more personalized and effective rehabilitation solutions.
Quan Liu, Siyuan Wang, Chang Zhu et al.· IEEE Transactions on Automat...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
