A multi-UAV trajectory optimization methodology for complex enclosed environments
MetadataShow full item record
Unmanned Aerial Systems (UAS) have become remarkably more popular over the past decade and demonstrate a continuous upward market trend. As UAS become more accessible and advanced, they are able to be incorporated into a broader range of applications and provide substantial operational benefits. In addition to exterior use cases, UAS are being investigated for interior use cases as well. An area that has great potential for UAV involvement are manufacturing and warehouse environments, as these typically occupy vast spaces. Warehouse logistics and operations are very complex and could significantly benefit from the integration of UAVs. Many companies are already exploring using UAS as a means to perform inventory audits to reduce labor costs and time, and improve accuracy and safety. To achieve the maximum benefit from this technology in these environments, multiple vehicles would be essential. The purpose of this thesis is to optimize the operations of multiple UAVs in complex and confined environments, using a warehouse model as a test case. There are added complexities when working with multiple vehicles; for example, ensuring that there are no collisions between vehicles. A great deal of research has been done on vehicle routing and trajectory optimization, but very little has been done with UAV optimization in confined spaces. This thesis further develops these algorithms and focuses in on the impact UAV involvement could have on operations in environments that are similar to warehouses. The proposed improvements from the current methods will help uncover the most optimal results by changing the process for finding solutions, the criteria under which solutions are ranked, and the operational/experimental setup. The new methodologies seek to resolve the sub-optimality issues from the existing approach to significantly reduce the mission time required to perform a warehouse inventory audit. An existing inventory scanning algorithm generates sub-optimal, collision free paths for multi-UAV operations, which has two sequential processes: solving a vehicle routing problem and determining optimal deployment time without any collisions. To improve the sub-optimal results, this thesis introduces three possible improvements on the multi-UAV inventory tracking scenario. First, a new algorithm logic which seeks to minimize the total mission time once collision avoidance has been ensured rather than having separate processes. Next, an objective function that seeks to minimize the maximum UAV mission time rather than minimizing the total of all UAV mission times. Last, an operational setup consisting of multiple deployment locations instead of only one. These proposed improvements are assessed based on their degree of impact on the overall mission time compared to the current methods. They are also analyzed in comparison to one another and in combination with one another to better understand the effectiveness and sensitivities of the presented changes.