AbstractIn order for an agent or a group of agents to achieve a goal, there is a sequence of actions that need to be taken. This sequence of actions constitutes a plan. In some occasions multiple plans that can be used to achieve the goal may exist. It may be the case that one wants to prevent an agent or group of agents from achieving a particular goal; currently, there is little research addressing this issue. Such research is valuable when preventing malicious agents or groups of agents such as bot nets from perpetrating destructive behavior.
In this thesis, we tackle the disruption of single agents and groups of agents by proposing a way to disrupt their plans to achieve goals. We describe a model of all possible plans that achieve the goal we wish to prevent, we provide algorithms that identify which parts of the plans should be disrupted in order to prevent achievement of the goal and, we also provide a way of disrupting these parts using norms. The focus of this thesis is summarized as follows.
1. Identify which parts of plans should be sabotaged in order to prevent achievement of a goal by some set of agents. In order to address this, we provide:
(a) a model that allows the representation of all possible plans that can be used to achieve a particular goal,
(b) a metric to measure disruption of plans and
(c) algorithms that identify which parts of plans should be disrupted in order to prevent achievement of some goal.
2. Identify which parts of plans should be sabotaged in order to prevent achievement of a particular goal by some set of agents, with as little impact as possible on other agents that are not part of the set (these agents are called external agents). In order to address this, we provide:
(a) an extension to the model from 1(a) to allow the representation of external agents' plans,
(b) a metric to measure disruption of external agents' plans and
(c) algorithms that identify parts of plans to be disrupted in order to render the desired goal unachievable while causing as little disruption to external agents' plans as possible.
3. Investigate the use of norms as a means to disrupt parts of plans using: (a) a model of agents that can keep track of norms and their plans to achieve a goal and (b) an algorithm to derive norms that disrupt the parts of their plans identified by the algorithms from 1(c) and 2(c) and so ensure that the goal cannot be achieved in specific situations.
We conclude with a summary of how disruption of agents' plans has been approached, high- lighting questions that remain open for further research. 3
|Date of Award||2016|
|Supervisor||Elizabeth Black (Supervisor) & Michael Luck (Supervisor)|