next up previous contents
Next: Limitations Up: Alternate Applications Previous: Intra-Modal Learning

Non-Human Training Data

One other note is that the systems described here are not specific to human interaction (although it is a compelling task). The interaction of other processes is also of interest. This includes the behaviour of animals ranging from the simple to the complex. One could consider learning the interaction of a cat and mouse as they run around the floor. If a simple overhead vision algorithm could determine their coarse locations, it would be possible to analyze some of their interactive dynamics and predict some trajectories. These concepts could be extended to the interaction of any coupled phenomenon provided a good representation appropriate for the ARL system is available. This could include two airplanes maneuvering in a dog-fight. Given their relative positions, it would be possible to compute non-linear trajectories and simulate the evolution of the engagement (if training data from previous dog-fights is learned).

Tony Jebara