A Novel Fall Detection System Using the AI-Enabled EUREKA Humanoid Robot

As a result of frailty in old age, falls are a leading cause of serious injury and death in older populations. As with most traumatic injuries, rapid treatment has the potential to dramatically improve outcomes; however, falls often occur in isolated areas, away from family and carers, preventing timely detection and response. Building on prior work by the authors using humanoid robots to enhance elder care in residential and hospital settings, this study investigates the potential to use robotic systems to detect and respond to falls. In this paper we describe a novel prototype system that integrates an intelligent humanoid robot, Robot EUREKA, with a machine learning NN classifier trained on sensor data from an associated mobile device, to accurately and reliably detect falls. This detection is used to direct the humanoid robot in real-time to respond to the casualty, with the capability to instantly alert carers and provide real-time patient monitoring and interaction through the robot. Following initial proof-of-concept success in the prototype, ongoing work seeks to extend the capabilities of Robot EUREKA to include embedded robotic vision, via neural image captioning, to (i) assess the type and severity of fall injuries to aid response and (ii) detect fall risks with a view to fall prevention. Neural image captioning for fall prevention and management will be piloted in the partnering ALTY Hospital in 2023–2024.