Activity Monitoring and Prediction for Humans and NAO Humanoid Robots Using Wearable Sensors

While humans or biped humanoid robots perform activities such as jogging and running, an accident event such as a fall may occur. This might cause damage to the human body or to the structural components of the robot. For humans, immediate identification of a fall will allow fast responses, while for a robot, early prediction can be used to take corrective measures to prevent a fall. Modern wireless sensing devices can be attached to humans or robots to collect motion data. We propose: 1) methods to learn and predict different activities for humans and robots; and 2) software tools to realize these functions on embedded devices. Our contributions include: 1) detection of falls for both humans and robots within a unified framework; and 2) a novel software development environment for embedded systems. Our empirical evaluations demonstrate that with high accuracy different types of fall-events are predicted using the same learning algorithms for humans and biped humanoid robots.