Navigation of Partially Observable Quadrotor in Smart Cities Based on UCT
Mohamed Farghal, Ayman El-Badawy · 2023
This paper focuses on the use of quadrotors in various applications within smart cities, such as surveillance, security, delivery, logistics, and emergency response. However, the complexity of navigating these environments is increased due to noisy or partial observations of the state, leading to the need for the quadrotor to make decisions based on incomplete information. To address this challenge, the paper employs the Partially Observable Markov Decision Processes (POMDPs) framework. By incorporating POMDP-based techniques, quadrotors can effectively manage limited information and make intelligent decisions. This implementation focuses on utilizing the Partially Observable Upper Confidence Tree (PO-UCT) algorithm, which acts as a planner within the POMDP framework. The PO-UCT method combines the UCT algorithm with belief state information, allowing the quadrotor to construct a belief about the underlying system state based on previous observations and actions. This algorithm is designed to find optimal strategies for decision-making in complex environments. Results show that the PO-UCT enables safe navigation and shortest path selection for the quadrotor in uncertain environments.