As part of a global initiative, innovators are working together to revolutionize the way a mechanical man’s most trusted companion engages with its owner, combining the powers of artificial intelligence (AI) and edge computing in a groundbreaking approach known as edge intelligence.
The venture is supported by a one-year seed grant from the Institute for Future Applied Sciences, a collaborative endeavour between the New Jersey Institute of Technology and Ben-Gurion University of the Negev.
Assistant Professor Kasthuri Jayarajah at NJIT’s Ying Wu Faculty of Computing is developing methods to design a socially assistive mannequin for her Unitree Go2 robotic canine that dynamically adapts its behavior and interaction style based on the characteristics of individuals with whom it interacts?
The primary objective is to bring a canine’s emotional and physiological responses to the forefront through innovative wearable technology, capturing vital signs such as introspection, emotional shifts, pain, and pleasure levels.
The invention is poised to make a profound impact on both dwelling and healthcare settings, providing a vital tool in combating loneliness among the elderly population and serving as a valuable aid in treatment and rehabilitation processes. Researchers led by Jayarajah are set to unveil groundbreaking findings on how robotic canines can effectively recognize and respond to non-verbal gestures from humans at the prestigious International Conference on Intelligent Robots and Systems (IROS) in 2023.
As principal co-investigator, Shelly Levy-Tzedek, an esteemed affiliate professor within the Division of Physiological Medicine at Ben-Gurion University (BGU), is a renowned expert in rehabilitation robotics, focusing research on understanding how aging and illness impact bodily function.
Researchers acknowledge the growing accessibility of wearable devices, noting that everyday items like earphones can be repurposed to monitor wearers’ physiological states, such as mental activity and micro-expressions? The venture aims to integrate innovative multimodal wearable sensors with traditional robotic sensors, for instance Utilizing a combination of visible and audio data, we can design systems that objectively and passively capture individual attributes.
While the notion of socially assistive robots is indeed captivating, a persistent obstacle to their widespread adoption lies in the dual challenges of affordability and scalability. Robots, such as the Unitree Go2, are unlikely to be equipped with the capabilities to handle giant AI tasks. Researchers have deliberately constrained processing power in comparison to massive GPU clusters, rather than simply limiting memory or shortening battery life, she noted.
To establish a solid foundation for the venture, initial efforts focus on building upon established sensor fusion techniques, while also delving into meticulously crafted deep-learning frameworks capable of enabling the development of commodity wearables that can accurately extract user characteristics and provide tailored movement guidance.
