Kress-Gazit, a professor of mechanical and aerospace at the Cornell University has developed algorithms and software called Linear Temporal Logic Mission Planning (LTLMoP) which has induced English understanding skills in the Humanoid robot Mae.
Kress-Gazit explained that the ultimate aim was to make the robot capable of understanding commands from anyone and respond to them.
The LTLMoP toolkit is a combination of language, control algorithms and logic. Mae is a 2 ft NAO humanoid robot developed at Aldebaran Robotics for performing tasks such as finding missing items at a grocery store. The robot is programmed to avoid spills that may be found at aisles and responds to specific commands that were issued. A controller required for such sophisticated tasks will need specific programming for enabling the robot to react to each of the possible situations that it may encounter. Such programming techniques are tedious to work on and likely to end up with errors or fail to work in certain situations. These techniques do not guarantee that desired response from the robot when it encounters certain situations. In case of LTLMoP it is possible to define high level specification for tasks in simple, structured English. The commands are written such that the robot can easily comprehend them, like conditional if…then statements or prepositional statements using ‘between’ etc. For instance, it is possible to tell Mae to visit the store corners and to look on either side while walking through the aisles.
Cameron Finucane, a graduate student said that instead of providing the robot with an ordered list of things, one can specify the kind of behaviour required by the robot.