The programming component is taking a modular approach to the problem. Each person is isolating a particular task and trying to tackle that problem.

 

1)      Matt Elliott: Take stemmatized sentences to tell the robot to do something. Trying to get the computer to understand stemma. Two inputs: stemma and the world around it.

a.     Takes the world in, processes, output can be a stemma. “Ok I’m here, this is what’s near me” Or this output says move towards this object, move towards this label.

2)   Jim Suffolk: Knowledge representation. Categorization Module. Action Selection. Creativity.

a.     Planning mechanism - Uses Temporal Difference Learning TD(Lamda) algorithm for high level planning in an arbitrarily defined state space that guarantees convergence on the optimal action. State space formulated in terms of categories and actions eligible to be taken from a particular state.

b.     Categorization - Determines existence of categories, planning algorithm can take actions to create new categories from what it sees in the world.

c.     Creativity - Formulation of new actions and categories. Re-representing the planning state space in terms of newly created actions and categories. Manages several separate representations of the environment in a manner similar to a genetic-algorithm in hopes that independently developed representations can be combined and mutated to give a very effective world representation.

3)    Kevin Jacobson: Given the world of matt, define what can be manipulated. Abstract classes based on matt’s world. From third to first person. Attentional aspect of the robot.

a.     Based on Jim’s module of categorization and mood

b.     Execution of Jim’s chosen Motor program

c.     Update matt’s world based on how he can manipulate the world

d.     Given the shape of the object

                                                               i.      Can be manipulated in some way

4)     Steve Dee: describe diagrams in a computer, with cognitive awareness. Let computers understand diagrams as well as we do. Cognitive Diagrammatics.

5)      Tyler Goeringer: Working on his own for the moment, but he is working on genetic algorithms.

a.     Computer learning