Search for: All records

Predicting flight trajectories is a research area that holds significant merit. In this paper, we propose a data-driven learning framework that leverages the predictive and feature extraction capabilities of the mixture models and seq2seq-based neural networks while addressing prevalent challenges caused by error propagation and dimensionality reduction. After training with this framework, the learned model can improve long-step prediction accuracy significantly given the past trajectories and the context information. The accuracy and effectiveness of the approach are evaluated by comparing the predicted trajectories with the ground truth. The results indicate that the proposed method has outperformed the state-of-the-art predicting methods on a terminal airspace flight trajectory dataset. The trajectories generated by the proposed method have a higher temporal resolution (1 time step per second vs 0.1 time step per second) and are closer to the ground truth. 

Unmanned aerial vehicles or drones are widely used or proposed to carry out various tasks in low-altitude airspace. To safely integrate drone traffic into congested airspace, the current concept of operations for drone traffic management will reserve a static traffic volume for the whole planned trajectory, which is safe but inefficient. In this paper, we propose a dynamic traffic volume reservation method for the drone traffic management system based on a multiscale A* algorithm. The planning airspace is represented as a multiresolution grid world, where the resolution will be coarse for the area on the far side. Therefore, each drone only needs to reserve a temporary traffic volume along the finest flight path in its local area, which helps release the airspace back to others. Moreover, the multiscale A* can run nearly in real-time due to a much smaller search space, which enables dynamically rolling planning to consider updated information. To handle the infeasible corner cases of the multiscale algorithm, a hybrid strategy is further developed, which can maintain a similar optimal level to the classic A* algorithm while still running nearly in real-time. The presented numerical results support the advantages of the proposed approach.