Decomposition methods for moving target search
Dissertation, Universität der Bundeswehr München, 2017
Gespeichert in:
| 1. Verfasser: | |
|---|---|
| Körperschaft: | |
| Format: | UnknownFormat |
| Sprache: | ger |
| Veröffentlicht: |
Neubiberg
Universität der Bundeswehr München
2017
|
| Schlagworte: | |
| Online Zugang: | Inhaltsverzeichnis |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Tag hinzufügen | Zusammenfassung: | Dissertation, Universität der Bundeswehr München, 2017 Airborne search and rescue missions are of incredible importance to save the lives of missing persons. Such missions must be planned carefully in order to optimize the chances of survival. However, planning must be conducted within a small time frame so as net to waste time. The target of the search is often likely to move, which significantly complicates the problem. The reason for the increased complexity is that the search reward at time k does not only depend on the observation made at time k, but on all observations made up until time k. In other words, the rewards over time are inseparable and, hence, state-of-the-art shortest path planning algorithms become inapplicable. A pilot who is faced with such a complex task in a stressful situation is prone to planning a suboptima] search trajectory. Coordinating a team of cooperating aerial platforms is especially difficult. Automation of search trajectory optimization is therefore the aim of this dissertation. Three novel problems are considered in this thesis: single platform search under kinematical constraints, single platform search under kinematical and resource constraints and strategy optimization for a team of heterogeneous cooperating platforms with shared resources. A mixed integer linear problem (MILP) formulation is proposed as well as a decomposition method for solving each problem. Computational experiments and simulations show that each proposed model and algorithm is applicable and efficient for solving its considered problem. The first problem variation is solved much faster using the proposed generalization of a oranch & bound algorithm compared to solving the MILP … |
|---|---|
| Beschreibung: | 157 Seiten Diagramme |