droftware
commented
7 years ago Two things need to be explored in order to tackle the above problem.
-
Upper Confidence Bounds These might be used along with the result given by bayesian networks(UCBB - Upper Confidence Bounds applied on Bayesian Networks) so as to increase the exploration of the agent.
-
Pheromones A time stamp for each explored micro-cell which can be used to prevent going to the same cells repeatedly.
An agent might get trapped in a local minima or might keep on visiting the same places again and again. While this might be good from an exploitative point of view, since its going to positions which ensure its greedy objective, however the agents totally lack an exploratory point of view since the agents do not explore the other remaining points which might be better.