With autonomous robots and small sensors beginning to turn into commercial commodities, the interest of computer scientists has shifted towards distributed computing. Distributed devices need to modify their behaviour in order to suit the altering situations.
However, their scope for understanding a circumstance is limited to local observations and behaviour exhibited for a certain case may pose a disaster for another. The programmers are therefore, faced with the daunting task of framing behavioural policies that can be commonly applied to all circumstances.
In order to come up with such behavioural policies, one would have to weigh all possible options for each device under all possible circumstances. Frans Oliehoek, a postdoc in Computer Science and Artificial Intelligence lab at MIT, is working on developing techniques to calculate such policies. The policy must take into consideration the device’s history while framing the calculation, which makes the policy framing even more complex. Oliehoek along with some colleagues has presented papers that describe numerous ways to bring down the scale of policy-calculation. Oliehoek claims that they have devised methods that could be implemented on distributed devices.
The important point is to identify the scenarios, in which the structural features of the issue imply that certain policy combinations don’t need separate evaluation. In some other situations histories can be merged with one another, but still point to the same result caused by the same action. Francisco Melo, an assistant professor at the Universidade Tecnica de Lisboa, claims that Oliehoek’s model is a generic one and can be applied to solving all kinds of decision problems. Melo added that the model is a complex one and hence cannot be applied for extremely small problems.
Source: http://web.mit.edu/research/