Agenda

Speaker:
Dr. Muhammad Rameez Ur Rahman, DAIS

Abstract:
Falls and the fear of falling severely impact the independence and mental health of the elderly. It is estimated that at least 1/3 of adults aged 65+ experience at least one fall per year. A common solution is to employ walkers and, more recently, smart walkers. Currently, smart walkers are costly, often lack the reliability needed to foster user trust, and are scarcely personalizable. The EasyWalk project addresses these challenges by developing a reliable, assistive device that leverages shared intelligence to safely navigate complex environments and restore user autonomy. In this talk, we focus on the core perception systems that enable the walker to robustly detect targets and obstacles. We introduce OpenNav, a zero-shot 3D object detection pipeline for real-time, open-vocabulary recognition of user-specified targets and unknown hazards without requiring prior 3D training data. To guarantee safe and socially acceptable navigation, the framework extends beyond static detection to include predictive spatial reasoning. Specifically, we present ECAM, a contrastive learning approach for human trajectory forecasting that minimizes environmental collisions, alongside a group detection module to identify social dynamics.

Bio sketch:
Muhammad Rameez Ur Rahman is a Postdoctoral Research Associate in Computer Vision and Machine Learning at Ca’ Foscari University of Venice, Italy, working on the PRIN project, EasyWalk, on intelligent assistive robotics. He received his Ph.D. in Computer Science from Sapienza University of Rome in 2024. His research focuses on computer vision, multimodal learning, and vision-language models, with applications in human behavior understanding, robotics, and autonomous systems.