Saturday, August 1, 2015
Training Asa H in real and virtual worlds
After each training session we typically record Asa H's hierarchical memory ( see, for example, blogs of 10 Feb. 2011 and 26 Aug. 2013). This can then be reloaded to begin a new training session if we so choose. Some training can be in real worlds involving things like Lego NXT robots and sensors. Other training can involve simulated robots in virtual worlds. It is quite possible to alternate between real worlds and simulations (virtual learning environments) as the memory is grown but one must be careful to avoid such things as sensory mismatches between the two sorts of robots. Adjusting the time scale of each of the training sessions is also tricky.