Robots that Can Guess Your Next Move Set to Smooth Our Everyday Lives?

By June 12, 2013

Man-machine interaction has now reached a new level with a robot that is programmed to anticipate people’s movements in advance, the better to help them with daily tasks.

Researchers are constantly endeavouring to optimise the ability of robots to mimic human actions with the aim of making them more useful for the general public. Following the example of a team at theNorwegian University of Science and Technologywho recently developed a robot capable of imitating human beings, researchers at Cornell University in New York State have now managed to programme a robot to predict a human being’s actions and so help him/her with the task in hand. Using a sensor, the robot observes the person’s movements and compares them with a database of videos before deciding just what it ought to do to help the person carry out the intended task. The researchers claim that this more intuitive approach will make the interaction between man and robot much smoother.

Identifying and anticipating gestures more accurately

The way this works is that the researchers give the robots access to a database of over 120 3D videos showing humans making everyday gestures and movements. The robot, which incorporates a Microsoft Kinect 3D camera, therefore learns to analyse the movements of the person in front of it, and then classify the actions into a predefined category – reaching for something, carrying, pouring, drinking and so on. This analysis also factors in details such as where the action takes place and the location of nearby objects, thus enabling the robot to make a best guess at the ultimate objective of a series of human actions. The researchers estimate that the robot predicts correctly in 82% of cases an action which takes place one second later. The robot is even more accurate and confident in identifying an ongoing gesture when the anticipation algorithm works in parallel.

Fields of application

This approach is likely to have useful applications in a variety of fields. For example, it could improve the quality of robotic telepresence (a system to facilitate remote communication and action using a camera and a screen to transmit a person’s face and voice). In addition, robots capable of anticipating human actions could be assigned to various workplaces to help people with their tasks – factories and hospitals being prime locations for this kind of support. Last but not least, older people could certainly benefit from technology that will help them to carry out household chores, thus increasing their independence from human carers.

Legal mentions © L’Atelier BNP Paribas