Accurate prediction of multi-agent behavior in dense and dynamic environments remains a central challenge for autonomous systems, with applications in robotic navigation, traffic management, and human-robot interaction. Conventional approaches often fail to model both the spatial interdependencies among agents and the long-term temporal patterns required for reliable trajectory prediction. To address this limitation, a multimodal framework is proposed that integrates Graph Neural Networks (GNNs), Long Short-Term Memory (LSTM) networks, and Inverse Reinforcement Learning (IRL). Spatial dependencies are represented through GNN layers, temporal dynamics are encoded by LSTMs, and reward functions are inferred from demonstrations via IRL. Multimodal fusion between trajectory histories and aerial imagery is further introduced to enhance environmental awareness. The framework is evaluated on the Stanford Drone Dataset (SDD) and SUMO simulations, where higher accuracy, robustness, and safety are observed relative to established baselines, including LSTM, Social-LSTM, and representative spatio-temporal models. A deployment-oriented analysis covering computational efficiency, robustness to noise/occlusion, and safety metrics (collision rate and time-to-collision) is also provided, underscoring the practicality of the approach for real-world navigation.

Global Positioning System (GPS) technology has significantly enhanced logistic activities in recent years by the precise and realtime monitoring of vehicle locations. Along with real-time location monitoring, gaining more informative data about vehicle activity has become increasingly vital for additional optimization. This is particularly the case for large-scale production processes such as street paving, where multiple vehicles are employed and the highly accurate coordination of their activities significantly impacts overall productivity and quality. While previous studies have predominantly focused on truck stop classification and delivery modes, this study suggests a novel approach to segment and classify vehicle activity along the entire delivery process into distinct operational phases: loading, traveling, waiting to unload, and unloading. A real-world GPS trajectory dataset of 165 transporters and 6,144 delivery trips was utilized to generate a feature-engineered dataset through rule-based preprocessing steps. Diverse machine learning models were fine-tuned and compared based on accuracy and F 1 Score to ascertain classification performance. The highest accuracy of 98% was achieved with the Random Forest (RF) classifier, demonstrating the potential of the model for accurately and automatically estimating vehicle states from available GPS data. The findings demonstrate the effectiveness of data-driven approaches in streamlining feet management and operational efficiency.

Wildfires not only pose a significant threat to human life and property but also have far-reaching impacts on communities and ecosystems. Effective prevention and mitigation strategies rely on accurate prediction of the path of these fires. This paper proposes the utilization of data obtained from Unmanned Aerial Vehicles (UAVs) to develop predictive models for fire spread. A comprehensive dataset is presented that includes key environmental variables that have been meticulously captured using these advanced technologies. The dataset comprises images from which essential features for predicting fire spread have been extracted. The method detailed in this article has been used to identify and incorporate crucial factors such as plant density, wind direction and speed, humidity, and geographical features. These key factors are then used to predict the spread of fires using Machine Learning (ML) techniques. After thorough study and comparison, AdaBoost and Random Forest (RF) demonstrate superior predictive capabilities. Evaluation metrics such as Mean Absolute Error (MAE) and Mean Squared Error (MSE) confirm the high accuracy and reliability of the proposed approach, achieving R-squared ($\mathrm{R}^{2}$) values above 0.98. By combining advanced technological tools with analytical methodologies, this approach has the potential to enhance fire suppression and management, safeguarding lives and assets.

Trust and security management in modern Intelligent Transport System (ITS) is a demanding task. In a recent publication a Large Scale Multimodal Data Processing Middleware for Intelligent Transport Systems (LDPM) was introduced. This LDPM depicts an ITS that utilises cryptographic and trust based technologies to provide secure Vehicular Ad-Hoc Network (VANET) communication. However, some critical aspects regarding evidence evaluation and trust scoring were subjected to future work. These topics are now addressed in this paper. Thus, a novel scheme to describe traffic related evidence in a multimodal environment, a modified version of a Bayesian Inference (BI) function, and a comprehensive data centric trust management method is presented. These findings integrate into the LDPM, but are also applicable in a stand alone solution. These goals were accomplished, as demonstrated in the final performance evaluation.

We present a first method for the automated age estimation of buildings from unconstrained photographs. To this end, we propose a two-stage approach that firstly learns characteristic visual patterns for different building epochs at patch-level and then globally aggregates patch-level age estimates over the building. We compile evaluation datasets from different sources and perform an detailed evaluation of our approach, its sensitivity to parameters, and the capabilities of the employed deep networks to learn characteristic visual age-related patterns. Results show that our approach is able to estimate building age at a surprisingly high level that even outperforms human evaluators and thereby sets a new performance baseline. This work represents a first step towards the automated assessment of building parameters for automated price prediction.

In this paper, a dataset of geotagged photos on a world-wide scale is presented. The dataset contains a sample of more than 14 million geotagged photos crawled from Flickr with the corresponding metadata. To guarantee the spatial representativeness of the dataset, a crawling approach based on the small-world phenomena and the Flickr friendship's graph is applied. Furthermore, the noisiness of user-provided tags is reduced through an automatic tag cleaning approach. To enable efficient retrieval, photos in the dataset are indexed based on their location information using quad-tree data structure. The dataset can assists different applications, especially, search-based automatic image annotation and reverse geotagging.

A huge number of user-tagged images are daily uploaded to the web. Recently, a growing number of those images are also geotagged. These provide new opportunities for solutions to automatically tag images so that efficient image management and retrieval can be achieved. In this paper an automatic image annotation approach is proposed. It is based on a statistical model that combines two different kinds of information: high level information represented by user tags of images captured in the same location as a new unlabeled image (input image); and low level information represented by the visual similarity between the input image and the collection of geographically similar images. To maximize the number of images that are visually similar to the input image, an iterative visual matching approach is proposed and evaluated. The results show that a significant recall improvement can be achieved with an increasing number of iterations. The quality of the recommended tags has also been evaluated and an overall good performance has been observed.

Data is the basis of the modern information society. However, recent natural catastrophes have shown that it is not possible to definitively secure a data storage location. Even if the storage location is not destroyed itself the access may quickly become impossible, due to the breakdown of connections or power supply. However, this rarely happens without any warning. While floods have hours or days of warning time, tsunamis usually leave only minutes for reaction and for earthquakes there are only seconds. In such situations, timely evacuation of important data is the key challenge. Consequently, the focus lies on minimizing the time to move away all data from the storage location whereas the actual time to arrival remains less (but still) important. This demonstration presents the dynamic fast send protocol (DFSP), a new bulk data transfer protocol. It employs striping to dynamic intermediate nodes in order to minimize sending time and to utilize the sender's resources to a high extent.

This paper introduces a new approach of approximating the selectivity of multimedia range queries. Estimating the selectivity of a range query is a pre-requisite to optimize a multimedia database query. We use the DBSCAN clustering technique for finding high density areas in the data set. Then, the selectivity is approximated with the help of a density function in combination with the volume of the query's hyper sphere. Our approach is fast and accurate which was evaluated on an image data set using the MPEG-7 scalable color descriptor. The technique is integrated with the help of the extensible optimizer architecture in the Oracle multimedia database system.

No abstract available.

No abstract available.