Multicriteria/multiobjective Decision Making in Humanitarian Operations - Path planning under Uncertainty

点击翻译按钮获取中文摘要

The act of searching is an important part of many humanitarian operations, such as search and rescue, demining, and of many surveillance operations for the purpose of protecting individuals, resources or infrastructures. Teams of searchers (humans and dogs) or autonomous unmanned vehicles (robots) may search for survivors, land or underwater mines, or illicit activities and abnormal behaviors. But, how and where to search? The answer lies in good and efficient search planning that ensures the best use of scarce and constrained search resources, while minimizing the risks to the search teams. Search agents must therefore define search areas and plan search paths in order to maximize the chances of their operations success, often in degraded and rapidly changing conditions. This is a particularly important phase, especially in the presence of uncertainty on the whereabouts, the detectability, and the conditions of the survivors, the threats, or the search objects. In this research program, we are interested in path planning under uncertainty in the presence of multiple and conflicting objectives. We strive to provide the planners (decision makers) with rigorous and efficient methods to help them address the challenges they face in uncertain and unfriendly environments. In order to do so, we intend to develop decision support tools based on rigorous mathematical models representative of the real-life problems encountered. This is needed in order to assist decision makers (search planners) in defining adequate and realistic search plans that take into account the multiple dimensions of search operations such as terrain constraints, physical constraints, visibility constraints, threat exposure, time constraints, detection capabilities, type of resources available, environmental conditions, etc. Research projects addressing path planning under uncertainty in the presence of multiple conflicting objectives are scarce. Most of the published results apply to situations where the goal is to get from a given point to a known destination, and are often based on a single objective optimization. Therefore, there is a need to advance research and ample opportunity for innovation in tackling this difficult problem. Our goal is to develop approaches integrating multiobjective optimization techniques that provide efficient solutions, with multicriteria decision analysis techniques that explicitly take into account the preferences and values of the decision maker in the loop. In addition, it is very important that decision makers understand how and why the proposed search plans were produced; these should not be perceived as the outcomes of a black box. For this reason, there is a great added benefit in providing the decision makers with explanation modules that present the rationale behind the proposed search plans. In this proposal, we will design, develop, validate and evaluate algorithms for path planning of multi-agents such as autonomous vehicles in the context of multiple objectives, with uncertainty on the whereabouts of the search objects, and on the detection capabilities. Subsequently, we will develop models and methods, based on discrete multicriteria decision analysis, that guide a decision maker’s choice of a path plan, that explicitly integrate his/her preferences and values, and that allow him/her to explore the various alternatives and visualize them with a geographic information system. Finally, we will explore how computational argumentation can be applied to provide explanations and argue for the proposed solutions. In addition to it being valuable for the decision maker, the introduction of computational argumentation in the multicriteria decision making framework would be leading edge research.