| Summary: | This research generates a large collection of optimized trajectories for multi-agent quadrotors. The hybridized algorithm extracts trajectories with various trade-off for all agents without discrimination. This allows the resources of all agents to contribute towards the completion of a task.
Two variations of multi-agent quadrotor missions are applied within this work. The first is spatially spread flight mission, MA-SPREAD whereas the second is formation flight, MA-FORMATION. The trajectories are designed within three environments: i) Highly Cluttered Indoor, ii) Cityscape and iii) Mountainous terrain. The initial path nodes are generated through a sampling based planner. Here, Rapidly Exploring Random Trees is expanded into Multi-Agent Rapidly Exploring Random Forest. These paths are used to form the initial population for Genetic Algorithm. Next, we apply Many-Objectives Optimization towards the optimization of all agents and its objectives.
This study strikes a balance between diverse and well minimized solutions through dimensionality reduction. Result shows that the algorithm can successfully find a diverse set of well minimized solutions within each environment. The end user will be supplied with high resolution visual imagery of each test environment and well-organized data that defines the trade-offs of each trajectory. These easy to understand information will assist the end user in making a final choice regarding the best multi-agent quadrotor trajectories for their mission.
|
|---|