AI, Robotics and Space

Browse

Browsing AI, Robotics and Space by Subject "11 Sustainable Cities and Communities"

Now showing 1 - 4 of 4
  • Thumbnail Image
    ItemOpen Access
    Automating emergency landing in cities with cognitive stress reducing explanations to pilot-in-the-loop
    (IEEE, 2024-10-29) Balakrishnan, Hamsa; Lechardoy, Bastien; Collignon, Maxime; Sen, Anwesha; Ali, Ahmed; Verma, Ankit; Wisniewski, Mariusz; Chatzithanos, Paris; Tsourdos, Antonios; Xing, Yang; Guo, Weisi
    Future smart cities will integrate urban air mobility (UAM) where electric vertical take-off and landing aircrafts (eVTOL) will improve labour and logistic mobility. One of the challenges in urban flight is emergency landing, where an eVTOL needs to land in an unofficial roof top or clear area safely within a time frame. In semi-autonomous eVTOLs, artificial intelligence (AI) is expected to assist in identifying emergency landing locations and navigation vectors, whilst the pilot flies and lands the eVTOL. Autonomously searching and correctly identifying safe candidate locations is important for safety of both the eVTOL and ground stakeholders. Furthermore, the processes of using AI to identify such locations should ideally be explainable to the pilot to reduce cognitive stress and aid the preparation of the emergency landing procedure. Here, we have developed a simulated emergency urban landing platform, whereby an eVTOL rotorcraft is scanning the ground for suitable emergency landing locations and actively explaining the navigation vectors to the pilot-in-the-loop. Cognitive stress is measured using real experiments using heart rate monitor, and an actor-critic deep reinforcement learning is used to learn what explanations are useful to reduce cognitive stress. The end result is that the eVTOL can identify emergency landing candidates, navigate to the safe landing zone (SLZ) whilst performing obstacle avoidance, and explain its decision making to the pilot-in-the-loop whilst minimizing its cognitive stress. We show through 3 scenarios that we can indeed reduce cognitive load significantly and also reduce the average time to reach the SLZ.
  • Thumbnail Image
    ItemOpen Access
    Drone pollution tracking in cities using recurrent proximal policy optimization learning
    (Institute of Electrical and Electronics Engineers (IEEE), 2024-10-29) Eliades, Andreas; Thellier, Elie; Tian, Haitao; Mehta, Shivam; Almesafri, Nouf; Chen, Hongqian; Wei, Zhuangkun; Perrusquía, Adolfo; Liu, Cunjia; Guo, Weisi
    Future smart cities will need to monitor polluters that periodically or burst emit illegal gases that is harmful to the environment. Tracking these sources in cities that have building obstacles and variation wind vector fields is challenging. Traditional methods using gradient kernels and partial-swarm-optimisation may not be suitable when the emissions are intermittent and pollution concentrations maybe trapped in local pockets. As such, step size tuning becomes difficult to generalise in these variational dynamic pollution environments. Here, in this paper, we have developed a simulated urban pollution propagation environment, whereby a drone is scanning the environment for gradients to search and localise the source. We consider both proximal policy optimisation (PPO)-based reinforcement learning and its recurrent PPO (R-PPO) alternative to achieve stable and reliable improvement of policy without the need to fine tune step sizes. We show localisation results across a range of wind, obstacle, and emission scenarios with success rate of 76-79% and high path efficiency of 95-96% in ideal conditions. When we examine alternative city structures and burst emissions, we can achieve success rate of 34% and path efficiency of 52%, showing that there is some generalisation in capability.
  • Thumbnail Image
    ItemOpen Access
    Intelligent 5G-aided UAV positioning in high-density environments using neural networks for NLOS mitigation
    (MDPI, 2025-06-01) Mousa, Morad; Al-Rubaye, Saba
    The accurate and reliable positioning of unmanned aerial vehicles (UAVs) in urban environments is crucial for urban air mobility (UAM) application, such as logistics, surveillance, and disaster management. However, global navigation satellite systems (GNSSs) often fail in densely populated areas due to signal reflections (multipath propagation) and obstructions non-line-of-sight (NLOS), causing significant positioning errors. To address this, we propose a machine learning (ML) framework that integrates 5G position reference signals (PRSs) to correct UAV position estimates. A dataset was generated using MATLAB’s UAV simulation environment, including estimated coordinates derived from 5G time of arrival (TOA) measurements and corresponding actual positions (ground truth). This dataset was used to train a fully connected feedforward neural network (FNN), which improves the positioning accuracy by learning patterns between predicted and actual coordinates. The model achieved significant accuracy improvements, with a mean absolute error (MAE) of 1.3 m in line-of-sight (LOS) conditions and 1.7 m in NLOS conditions, and a root mean squared error (RMSE) of approximately 2.3 m. The proposed framework enables real-time correction capabilities for dynamic UAV tracking systems, highlighting the potential of combining 5G positioning data with deep learning to enhance UAV navigation in urban settings. This study addresses the limitations of traditional GNSS-based methods in dense urban environments and offers a robust solution for future UAV advancements.
  • Thumbnail Image
    ItemOpen Access
    Machine Learning driven complex network analysis of transport systems
    (Elsevier, 2025-07) Xia, Yuqin; Wang, Kewei; Tanirat, Purin; Lee, Bryan; Moulitsas, Irene; Li, Jun
    A complex network is a system of interconnected nodes linked by edges, exhibiting non-trivial structural features such as community structure or scale-free distributions. This study develops a novel and generic Machine Learning-driven framework that integrates Complex Network Theory and Machine Learning methods for a comprehensive and multifaceted analysis of transport systems. Specifically, four key functional development and analysis are undertaken: 1) Network analysis, using complex network indicators to study the static properties of the transport systems; 2) Network clustering, employing K-means and hierarchical clustering methods to identify underlying community structures; 3) Network resilience, examining the networks' dynamic characteristics and structural evolution under escalating node attacks to evaluate their robustness; 4) Link and feature prediction, developing Graph Convolutional Networks (GCNs) and Multi-Layer Perceptron (MLP) models to predict hidden links and features. The proposed framework is subsequently applied to two distinct transport systems, namely, the China railway network and the Paris multi-modal transport system. The complex network analysis reveals distinct complex network features in network scale, density, and efficiency, yet both demonstrate a power-law distribution. The clustering analysis based on various node and edge properties exhibits a pattern of concentric circles, radiating outward from the urban to peripheral cities in China railway network, while a high density of short-distance connections within central Paris and a prevalence of long-distance connections in the outskirts. The network attack simulations show fine resilience of the Parisian multi-modal system and low resilience of the China railway network. For link prediction, an encoder-decoder model based on GCN and multiple MLPs are developed for various scenarios. The results for the China railway network reveal critical interregional links, emphasizing the need to strengthen regional connectivity, such as expanding the high-speed railway between Hainan Island and the mainland, and establishing a major transportation artery running from south to north. In the Paris transport system, this study predicts an interesting link extending from southern Paris eastward toward northern Seine-et-Marne, indicating a demand for a direct connection. For both networks, the hidden links are largely concentrated in more developed areas, likely driven by strong economic and social interaction demands, highlighting the need for more balanced transport network development. Overall, the results of this study align closely with existing literature and official transport development plans. This research contributes to the theoretical development in Complex Network Analysis using Machine Learning and offers valuable insight to improve the two transport systems.