Conflict-free urban air mobility planning with an airspace-resource-centric approach

• Main Author: Dai, Wei
• Other Authors: Lyu Chen
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering
• Online Access: https://hdl.handle.net/10356/174062
• Institution: Nanyang Technological University

Urban Air Mobility (UAM) has come to the sight of people as a new mobility type that has the potential to be a game changer in large cities' traffic systems. The conventional air transport system has been operating for over a hundred years with great performances in both efficiency and safety, which raises high requirements for the future operations of UAM. To enable the UAM to assess the urban airspace without impeding the operation of existing air traffic, research and developments are needed incorporating the specific features of UAM. Among the major services provided by the future UAM management system, strategic trajectory planning is of great importance as it contributes to both improving operational efficiency and flight safety. Existing trajectory planning methods for conventional aviation can hardly be applied to the context of UAM, and the studies in the literature on trajectory planning methods specifically addressing the challenges in UAM are limited. An advanced trajectory planning method for UAM is needed to incorporate the features and requirements of UAM traffic demands and eVTOL aircraft performances. Motivated by the need for a UAM trajectory planning method, this research unfolds in several stages. Firstly, we introduce the airspace-resource-centric (ARC) concept for urban airspace management. Whether assuming full implementation of the free-flight concept or adopting the air-route-based architecture will not satisfy the requirement of UAM. The ARC concept enables route-free flight planning with better flight predictability than the free-flight concept. It discretizes the spatial volume of urban airspace, and thus also discretizes the urban airspace resources. Dynamic assignment of airspace cells enables 4D management of urban airspace, including spatial volume, navigational services, etc., facilitating quantitative airspace utilization. The concept also leads to a graph-based airspace model for further usage. The second and third stages of this thesis aim to answer the questions supplementing the ARC concept. In the second stage, we analyze the traffic flow distribution based on the structure-free ARC concept. To achieve this goal, a 4D path planning method with strategic conflict avoidance is developed, namely the conflict-free A* algorithm. Numerical study results show that the algorithm can generate multiple flight paths without duplicated airspace cell utilization. The result also shows that the structure-free concept leads to an imbalanced vertical utilization of urban airspace. From the perspective of traffic management, a more balanced traffic flow is preferable. Thus a layered utilization of the airspace is recommended and will be used in the consequent stages. In the third stage, we analyze how the airspace cell shape affects flight safety. We studied the relationships between crossing traffic patterns, centralized tracking system performance, and conflict detection effectiveness provided by the tracking system. To achieve this probabilistic assessment, we propose a Monte Carlo simulation-based framework, which enables the evaluation of the likelihood of successful conflict detection and detection delay in a quantitative manner. The result suggests that orthogonal crossings are more likely to be detected, which supports the selection of block shape as the airspace cells. In the fourth stage, we develop a data-driven power consumption model for eVTOL aircraft. Precise estimation of aircraft power consumption with given kinematic states and environmental states is essential in the management of battery energy and prevents battery draining which will cause serious accidents. Motivated by the absence of a precise power consumption model that can be applied to multiple eVTOL aircraft types, we use the ensemble learning method to model the power consumption of eVTOL aircraft. The model that we develop outperforms the existing models in the literature. The model will be used in the evaluation of trajectory planning from the perspective of power consumption. Lastly, in the final stage, we present a multi-objective 4D trajectory optimization model using the ARC concept. We formulate the trajectory optimization problem mathematically. The formulation is a non-convex optimization problem with both continuous and 0-1 variables and implicit functions, which is difficult to resolve using mathematical programming or meta-heuristic optimization. To solve this problem, we proposed a method based on deep reinforcement learning (DRL). The trajectory optimization is reformulated as a Markov Decision Process (MDP), and a Parameterized Q-Network with Action Masking (PDQNAM) is developed for the discrete-continuous hybrid action space. To handle multiple objectives, a hierarchical approach is introduced, leading to efficient convergence of the model in simulation results.