This paper offers an improvement for machine vision based tracking algorithm using NARX neural network. This is achieved by predicting the next position of a mobile agent based on previous movement described by time-series functions. The algorithm remains passive and does not interfere with the quadrotr flight path until detection of multiple agent-candidates, when the tracking algorithm is not able to identify or locate the assigned agent. Meanwhile the NARX training set are updated in parallel with each successful tracking cycle. This process is done to eliminate the overfitting problem of BPTT training.
Keywords: NARX, quadrotor, agent tracking, local minimum problem, machine vision
Specialized robots are occupying a vast share of researches worldwide. In terms of aerial specialized robots, autonomy is the one of the major subjects that are being studied. As known, flight autonomy is consisting of three parts: Mission planning, trajectory generation and path tracking. While it exists variety of local and global generation algorithm, UAV path planning is always associated with machine vision. This intellectual approach is solid because it can give a dual supervision link to the situation being surveyed. The operator has always the authority to step in and stop any unwanted action. The machine vision based algorithm has a huge disadvantage due to homogeneity and identity factors. This is because from distance object tend to look the same geometrically as well as color wise. In this paper we offer a solution to tackle the homogeneity problem between tracked agents which can cause local minimum problems. The UAV will forecast the next position of the tracked agent using Artificial Neural Networks. Position estimation is based on registered previous coefficient.
Keywords: position prediction, path planning, quadrotor, neural networks