Project Abstract

Research on autonomous navigation of Unmanned Aerial Vehicle (UAV) plays a pivotal role in designing reliable industrial UAVs for diverse applications. In our MOSTI TeD1 project, we are delving into the great potential on autonomous navigation under tree canopy (equivalent to partially GPS-denied environment).

Such UAVs will be extremely valuable for security and data collection in Malaysian farms or plantations. Hence, we propose a Deep Reinforcement Learning (DRL) aided approach that enables the proposed UAV to execute successful autonomous navigation tasks under a tree canopy. The mission is formulated by exploiting the Partially Observable Markov Decision Process (POMDP) and solved by a novel DRL algorithm (in parallel, the conventional schemes will also be evaluated as benchmarks).

The project also provides our students to collaborate with researchers from the University of Southampton Malaysia, Aerodyne Group, and Universiti Sains Islam Malaysia to develop and implement the required algorithms for DRL-aided under-canopy navigation.

Furthermore, we will have a precious chance to comprehend situational awareness based on sensor integration and fusion, as well as how it can be exploited for designing our targeted autonomous navigation algorithms step-by-step. Throughout the research endeavours, we will learn how to design an entire system as well.