초록

Abstract ▼

연구개요
Unmanned aerial vehicles (UAVs), have impressive capabilities. However, single UAV has limited to conduct various missions. A system of UAVs and replenishment service stations (RSSs) can be used to address issues of social concern such as search & rescue, border patrol, etc. Allocating the s

연구개요
Unmanned aerial vehicles (UAVs), have impressive capabilities. However, single UAV has limited to conduct various missions. A system of UAVs and replenishment service stations (RSSs) can be used to address issues of social concern such as search & rescue, border patrol, etc. Allocating the system resources well directly relates to lives saved. As such, determining near optimal allocation of the robotic resources is an essential problem facing the use of UAVs for large-scale complex social applications. We call this the robotic allocation problem (RAP). The goal of the proposed research is to identify classes of robotic allocation problems (RAPs), develop formal models describing them, create efficient solution methodologies for them and create prototype systems. It is our ultimate goal to create next generation methods that will enable the UAV systems of the future to be realized for the benefit of our society.

연구 목표대비 연구결과
In this study, we present two RAP classes and present five tasks to achieve them.
Task 1 shows the formal definition of our RAP problem. We classified and modeled the generic RAP. Description of Task 1 is embedded of each task.
Task 2 shows the efficient methods for the allocation of resources in deterministic RAPs. We devlop MILP(Mixed Integer Linear Programming) model of deterministic RAP. To get exact solution within the reasonable time, we apply column generation method.
Task 3 shows the efficient methods for the allocation of resources in stochastic RAPs. We develop 2 types of model whether RSS is fixed or not. In fixed design case, we develop generic network model of stochastic RAP and automatically generate MDP (Markov decision process) model and simulator from any given network. To solve complex MDP model, we develop 2 heuristics. In unfixed design case, we jointly determine RSS and UAV’s task. We formalize MDP model for our system design problem. To solve the MDP model in complex case, we develop deep reinforcement learning based learning approaches.
Task 4 shows the develop resource design methodologies which allow the determination of an optimal number of resources for RAPs. We develop pertri-net model of UAV systems. To find the number of resources in complex case we develop saving heuristics and modified genetic algorithm for our RAPs.
Task 5 shows developing prototypes for systems of UAVs and RSSs in class 1 & 2. To verify the our operation methodologies in real environment, we develop prototypes of systems. We develop our own control packages of UAVs and develop replenishment service stations. In addition, we conduct UAV tests in real environment.
The current research contents are very close to the our existing goals. The contents of the study are judged to have achieved the our goal. The results of the study were summarized and faithfully reported and presented to domestic and international conferences. In addition, we are in process of preparing to submit several top-tier journals.

연구개발결과의 중요성
Given the rapidly increasing unmanned aerial vehicle market, it is no doubt that it will become a national future growth engine. However, in order for unmanned aerial vehicles to be used directly in real life, research on unmanned aerial vehicles system is essential and it will be competitive in the future to overcome some shortcomings arising from the nature of UAVs such as limitation of fuel and capacity. In this regard, our research on resource allocation of unmanned aerial vehicle system suggested the aspect that unmanned aerial vehicle system should have, and also determined how resources should be allocated and distributed in the example of use. In addition to theoretical research, we verified the usefulness of the proposed unmanned aerial vehicle system by confirming whether it is actually applicable through prototype development.

(출처 : 요약문 2p)

목차 Contents

• COVER ... 1
• 연구결과 요약문 ... 2
• 목차 ... 3
• 1. 연구개발과제의 개요 ... 4
• 2. 연구수행내용 및 연구결과 ... 4
• 2.1 Efficient methods for the allocation of resources in deterministic RAPs ... 4
• 2.2 Efficient methods for the allocation of resources in stochastic RAPs; ... 6
• 2.3 Develop resource design methodologies which allow the determination of an optimal number of resources for RAPs ... 12
• 2.4 Developing prototypes for systems of UAVs and RSSs in class 1 & 2 ... 15
• 2.5 Achievement of the project ... 17
• 3. 연구개발결과의 중요성 ... 17
• 4. 참고문헌 ... 19
• 5. 연구성과 ... 19
• 대표적 연구실적 ... 22
• End Of Page ... 32