Cycle counting with UAVs: Sample selection in time-restricted scenarios with neural network predictions

This thesis is part of a research for setting up an autonomous UAV solution for cycle counting. This research focused on the autonomous construction of cycle counting routes. The research proposes a method that focuses on cycle counts in time restricted-scenarios. A neural network is used which predicts the chance that a location's inventory record status is inaccurate. Th model's objective is to maximise the sum of the predicted chance of being inaccurate of locations visited each cycle count.