If you’re constantly second-guessing whether or not you turned the gas off, a new robot could be your best and newest automated companion.
A new robot called the Watch-Bot, which was presented this year at Sweden’s International Conference on Robotics and Automation, watches humans as they go about their daily lives at home so it can learn to point out when they forgot to do something.
The machine consists of a Microsoft Kinect mounted on a tripod, a laptop for its brain center, a movable head, and a laser pointer. If you were making coffee in the kitchen, for example, the 3D cameras in the Kinect would track the activity and use the laser pointer to highlight the carton of milk you forgot to put back in the fridge.
The Watch-Bot is like the passive aggressive roommate who leaves post-it notes on the fridge, except you won’t cause drama if you decide to scream at it.
According to the Watch-Bot proposal, the average human forgets three facts, chores, or events everyday, which makes it necessary for a robot to be able to track what the human forgot to do, in addition to what the human is doing. Before the Watch-Bot can remind you of something you forgot to do, it has to learn your gestures and activities. It uses the 3D scenes collected by the Kinect to establish the different steps in a particular activity and consequently infer intent so that it can pick up the slack of your forgetful mind.
Take the daily routine of pouring milk in your coffee, for instance. The Watch-Bot observes the sequence of actions as well as the objects the human interacts with to split the activity into four discrete parts: fetch-milk-from-fridge, pour, put-milk-back-to-fridge, leave. The system is modeled to temporally relate the segmented actions so it knows that pouring the milk follows removing the carton from the fridge. Relying on its learned sequence of actions, the Watch-Bot points out the milk carton since it has learned that the milk goes back in the fridge before you leave. Because the Watch-Bot is self-taught, its approach becomes more methodical and accurate over time.
While it would certainly be helpful to have all the things you forgot to do in a day pointed out to you by something without emotion, not all of us do everything the exact same way everyday. Maybe you left the room before putting the milk back in the fridge on purpose because you’re going to pour it in your cereal in a few seconds. Would the Watch-Bot be able to detect that? On the other hand, if the robot did learn to trace your actions to the tee, that would probably make someone painfully aware of the monotony of their life. When a robot can predict your next move, that might be the time to shake things up a bit.