The researchers from Stanford and Cornell introduced a project known as Watch-Bot at ICRA that can be an ideal supporter for all kinds of household activities. It delivers reminders of faulty activity patterns that may take place in performing the household activities. For instance, if you forget milk outside or leave food inside the microwave, then the robot will identify it and will gently remind you.

 

The watch-bot is incorporated with a 3D sensor, based on Kinect, a laptop, a laser pointer and a camera that can tilt and pan. The robot visited the office and kitchen and spent an entire week watching people go about doing their roles. It captured 458 videos of human activities, which were later on interpreted with 23 types of objects and 21 distinct actions and out of 458 videos almost 222 videos have results wherein a person forget one thing or the other.

 

 

Proper training is provided to the Watch-bot through videos of humans mostly memorizing to do stuff. However, it was not disclosed to the Robot the actions that were supposed to be performed, but that humans have forgotten to do. It identified on its own the activities that were needed to be performed with the aid of its probabilistic learning models potent enough to detect relations and patterns directly from the Kinect data and the camera. The AI researchers consider this approach as unsupervised learning and as an alarm reminder, the Watch Bot uses its laser pointer to aim the object as a reminder.

 

During the experiment process, Watch Bot was able to inform when humans forgot to perform a task and remind them about it for almost 60% of the time. Few of the tasks that were successfully reminded by the Bot were turning the screen off, placing the milk back into the refrigerator, taking food out from the microwave and similar. According to the researchers, even most of the participants in the study program confirmed that the robot was supportive.

 

Fundamentally speaking, the robot is not aware of the objects and materials that are used by humans in their homes on a daily basis. It performs its activities by learning algorithms that track problematic patterns for identifying forgotten actions. It is a plus point of the Watch Bot that makes it intelligent enough to adapt novel situation and identify patterns that can be supportive over time, especially when new settings have to be made. Presently, the researchers are working to enable the robot to enhance its learning from online videos and other user-generated content from social platforms, which is an element of the project known as RoboWatch.

 

Conclusion – As far as robots are concerned, Watch Bot is not precisely a spectator. It is not a finished product, but it is introduced as a proof of the underlying technology concept that can be inculcated in a wide range of robots as long as they are incorporated with an RGB-D sensor such as Kinect and more and also a laser weapon. Even if it does not deliver results always, it is still an item with features that are worth paying.