Delivery Drones Dispatch Policy Under Wind Uncertainty

Surveyed the state-of-the-art trajectory generation with obstacle problems for UAV, which can be solved by optimal control via sampling-based search, graph-search and artificial potential fields method.

Classified delivery tasks as high priority and low priority drones respectively to perform simultaneous exploration and exploitation in unknown environments.

Proposed exploration tasks aiming to reduce the quantified uncertainty, which implicitly aids task completion in the long term. While exploitation tasks making full use of gathered information to optimize operation.

Leveraged Gaussian Process Regression (GPR) to quantify wind uncertainty. Based on this, learned model is advantageous in estimating the confidence of proposed path and proposal of further exploration tasks.

Present Markov Decision Process (MDP) framework for Informative Path Planning (IPP) with mutual-information for both high priority and low priority delivery drones to better explore and exploit.