Algorithms that can make a collaborative robot arm react to moving objects within milliseconds could mean a future where robots and humans work closer together on production lines. Jim Drury reports.
Industrial robot arms are useful but limited, due to their lack of awareness of their surroundings. Norwegian researchers used this Kinect sensor and a unique series of algorithms to help one such arm adapt to moving objects within milliseconds. The aim is to prevent industrial accidents and allow humans to work in closer proximity to robot arms - which can weigh several tonnes. SOUNDBITE (English) MARIANNE BAKKEN, SINTEF RESEARCHER, SAYING: "We use a 3D sensor - microsoft Kinect - to generate a point cloud from the scene. The camera uses an infra-red pattern and a camera to make this point cloud, and from the point cloud we get a 3D map of the space around the robot like here. So when a human comes into the scene you can see that coloured. So we know where the obstacles are, and based on the 3D map we have made a really fast algorithm that computes a safe path for the robot." SOUNDBITE (English) ØYSTEIN HOV HOLHJEM, SINTEF RESEARCHER, SAYING: "I can just put my hand here and the robot will quickly react, and the robot is able to plan a new path from its current position, all the way to the goal, within less than 20 milliseconds." Collisions are a constant danger for existing industrial robots, because it takes them a matter of seconds to adapt to unpredictable movement. The researchers think their algorithms could revolutionise industrial production. SOUNDBITE (English) MARIANNE BAKKEN, SINTEF RESEARCHER, SAYING: "We believe that the algorithms we have developed here at SINTEF can enable human-robot collaboration in, for instance, manufacturing industry. This could be highly relevant for tasks such as pick and place and assembly, where robots could give human workers a helping hand and increase efficiency a lot. And this could also be very useful for smaller companies where you could use collaborative robots for many different tasks where you usually have no automation today." The team plans to sell the concept to the industrial sector.