As Unmanned Aerial Vehicles (UAVs) are becoming crucial in modern warfare, research on autonomous path planning is becoming increasingly important. The conflicting nature of the optimization objectives characterizes path planning as a multi-objective optimization problem. Current research has predominantly focused on developing new optimization algorithms. Although being able to find the mathematical optimum is important, one also needs to ensure this optimum aligns with the decision-maker’s (DM’s) most preferred solution (MPS). In particular, to align these, one needs to handle the DM’s preferences on the relative importance of each optimization objective. This paper provides a comprehensive overview of all preference handling techniques employed in the military UAV path planning literature over the last two decades. It shows that most of the literature handles preferences by the overly simplistic method of scalarization via weighted sum. Additionally, the current literature neglects to evaluate the performance (e.g., cognitive validity and modeling accuracy) of the chosen preference handling technique. To aid future researchers handle preferences, we discuss each employed preference handling technique, their implications, advantages, and disadvantages in detail. Finally, we identify several directions for future research, mainly related to aligning the mathematical optimum to the MPS.

As Unmanned Aerial Vehicles (UAVs) are becoming crucial in modern warfare, research on autonomous path planning is becoming increasingly important. The conflicting nature of the optimization objectives characterizes path planning as a multi-objective optimization problem. Current research has predominantly focused on developing new optimization algorithms. Although being able to find the mathematical optimum is important, one also needs to ensure this optimum aligns with the decision-maker's (DM's) most preferred solution (MPS). In particular, to align these, one needs to handle the DM's preferences on the relative importance of each optimization objective. This paper provides a comprehensive overview of all preference handling techniques employed in the military UAV path planning literature over the last two decades. It shows that most of the literature handles preferences by the overly simplistic method of scalarization via weighted sum. Additionally, the current literature neglects to evaluate the performance (e.g. cognitive validity and modeling accuracy) of the chosen preference handling technique. To aid future researchers handle preferences, we discuss each employed preference handling technique, their implications, advantages, and disadvantages in detail. Finally, we identify several directions for future research, mainly related to aligning the mathematical optimum to the MPS.

Information overload and the labor-intensive nature of the military decision-making process have made human decision-making deficient in modern warfare. To mitigate the current deficiencies, AI systems have been proposed. In a military context, these systems should be reliable and explainable, warranting a human-AI collaboration approach. One process that can greatly benefit from decision support is military helicopter route planning. It has many different conflicting criteria related to it, necessitating multi-objective optimization (MOO) techniques. In the literature, there is a lack of research on the decision-maker’s role. To fill this gap, interactive MOO is proposed to utilize the advantages of human-AI collaboration and create a shared understanding of the problem and its tradeoffs. Unlike the current interactive MOO research, the project is done with the active involvement of domain experts. Four different studies are proposed to investigate this problem. First, a literature study will be performed to review the state of the art. Next, an interactive MOO method will be designed, which will be applied to an Air Interdiction mission. Simultaneously, we will investigate the effect of the level of detail of threat modeling on the resulting optimization problem. Finally, the decision-support tool will be applied to a Strike Coordination and Reconnaissance mission.