Robot learning through social media crowdsourcing

Methods designed to enable robots to learn on their own is a heavily studied area. If robots are to become an integral part of our society, they must possess the ability to learn without direct guidance from a dedicated user. Robot owners will not enjoy the duty of teaching their robot everything it knows. The ability for a robot to utilize various resources in its environment will enable its learning capabilities to be self-guided and independent. This paper investigates the use of social media crowdsourcing to allow a robot to access the vast information gathering resources available on Twitter. Specifically, the robot will record a human performing simple physical actions, upload the video to its Twitter account, and ask its followers for a description of the actions. The recorded parameters of each action is utilized as input into a multi-class support vector machine (MC-SVM) classification algorithm, which will enable the robot to recognize the action at a future time.