To develop better traffic prediction models, it is crucial to consider the spatiotemporal characteristics of traffic network and the effects of many external factors such as weather, land use patterns etc. A knowledge graph stores relationships among these external factors by turning data into machine-understandable information. It can accurately represent the correlations between heterogeneous external data sources and transportation network. However, integrating knowledge graphs and traffic networks is challenging due to the inherent heterogeneity of information present in external factors. To address this issue, this study introduces a traffic prediction model fused with a knowledge representation technique. A knowledge representation learning model called 'TransE' has been used to produce knowledge graph embeddings which are added with traffic features as inputs to several machine learning-based traffic prediction models. The models are trained over traffic data collected in Florida's Seminole county from January to December 2019. Without knowledge graph embeddings, Random Forest model outperforms other prediction models with R2 value of 0.78 (RMSE=148.18). However, when knowledge graph embeddings are used as additional features, the R2 value of a Random Forest model increased to 0.85 (RMSE=120.67). Experimental results show that embedding knowledge graphs can significantly improve the accuracy of traffic prediction models.
We present a pilot study exploring the potential of Large Language Models (LLMs) to interface with application programming interfaces through logical instructions, specifically within the domain of Geographic Question Answering for route optimization. This study employs a Continuous Retrieval-Augmented Generation approach combined with fine-tuned LLMs, featuring customized node-based storage and vector search retrieval. We also provide a comparative analysis of the method's effectiveness and adaptability in handling diverse textual queries.
This paper introduces a novel approach to solving a practical variant of the Vehicle Routing Problem (VRP), the multi-trip VRP with heterogeneous fleet and time windows (MTVRPHFTW). The approach integrates an improved Ant Colony Optimization (IACO) metaheuristic, a modified density-based spatial clustering of application with noise (DBSCAN-Plus) clustering, and a Micro-Cluster Fusion Scheme. The proposed framework aims to optimize vehicle routes by minimizing total travelling distance and time while ensuring a fair distribution of workload among the vehicles (drivers). To evaluate the proposed algorithm, referred to as the Ant Colony Optimization (ACO) algorithm with improvement mechanisms (Cluster Improved Ant Colony Optimization, CIACO), real-world data from a logistics company in Canada was utilized. This empirical testing aims to validate the algorithm's effectiveness in practical applications. The experimental results of CIACO demonstrate that the proposed algorithm outperforms existing methods in terms of reducing traveling distance, minimizing traveling time, optimizing the use of smaller vehicles to reduce CO2 emissions, achieving balanced workloads among drivers, and improving overall route optimization.
Given an origin, a destination, and a directed graph in which each edge is associated with a pair of non-negative costs, the bi-objective routing problem aims to find the set of all Pareto-optimal paths. This problem is societally important due to several applications, such as route finding that considers both vehicle travel time and energy consumption. The problem is challenging due to the potentially large number of candidate Pareto-optimal paths to be enumerated during the search, making existing compute-on-demand methods inefficient due to their high time complexity. One way forward is the introduction of precomputation algorithms. However, the large size of the Pareto-optimal set makes it infeasible to precompute and store all-pair solutions. In addition, generalizing traditional single-objective hierarchical algorithms to bi-objective cases is nontrivial because of the non-comparability of candidate paths and the need to accommodate multiple Pareto-optimal paths for each node pair. To overcome these limitations, we propose Multi-Level Bi-Objective Routing (MBOR) algorithms using three novel ideas: boundary multigraph representation, Pareto frontier encoding, and two-dimensional cost-interval based pruning. Computational experiments using real road network data demonstrate that the proposed methods significantly outperform baseline methods in terms of online runtime and precomputation time.
Emergency Response Vehicles (ERVs), such as firetrucks, ambulances, etc. operate with the purpose of saving lives and mitigating property damage. As such, ERV travel-time reductions may result in significant benefits to the community. A common strategy to improve travel times is Emergency Vehicle Preemption (EVP). EVP seeks to reduce ERV delays by providing the right-of-way to ERVs as they approach an intersection. This study proposes a new Dynamic Preemption strategy that determines the need for preemption prior to the ERV reaching the vicinity of the intersection, utilizing real-time data streams. This paper evaluates the effectiveness of some existing and proposed preempt control strategies using a digital twin testbed consisting of a series of signalized intersections on an urban arterial in Georgia. The best EVP strategy maximizes the improvement in ERV travel time while minimizing the adverse effect of preemption on the traffic in conflicting directions. Therefore, this study evaluates both the positive impact of EVP on the ERV as well as the adverse impact on the cross-street traffic. The study found that the potential exists for significant improvements in ERV travel time with the proposed Dynamic Preemption strategy, with minimal impact to the conflicting traffic. For the simulation corridor there was a 20% reduction in the ERV travel times with the implementation of the Dynamic Preemption strategy, compared to traditional EVP practices.
