New Hampshire Department of Transportation

This paper will discuss how to develop Explainable and Trustworthy AI to decide on infrastructure in a data-driven way with a particular focus on Transportation Infrastructure and Highway Engineering. This paper suggests a Hybrid Explainable AI model combined with Federated Learning to overcome major challenges of Intelligent Transportation Systems (ITS) and Smart Infrastructure. The proposed approach, based on the use of TensorFlow Federated (TFF), can guarantee privacy of data by training models in a decentralized manner, as well as increase the level of decision transparency with the help of such tools as SHAP and LIME. This blend of high-tech approaches will lead to the development of Decision-Centric AI Systems that are not only maximizing traffic flow and infrastructure planning but also in line with the concepts of Responsible Infrastructure AI. This research has shown that the hybrid solution can enhance scalability, interpretability, and real-time decision-making performance, being a strong solution to the implementation of AI-driven smart infrastructure systems in the city and highway context.

Road networks fragment wildlife habitat and impede wildlife connectivity, which leads to elevated wildlife-vehicle collision (WVC) risk and increased danger to humans and wildlife. Habitat connectivity has been linked to WVC hotspot location and intensity, but this relationship likely depends on landscape context and road characteristics, which may be nonlinear due to varying habitat availability. Our objective was to evaluate factors affecting WVC location and intensity across New Hampshire, USA, with a focus on habitat connectivity. We assessed the relationship between WVCs and five connectivity models using generalized additive models and compared connectivity effects to road and land cover characteristics. We found that a barrier-sensitive wildlife species connectivity model was the best predictor of WVC hotspots and had a strong, negative nonlinear relationship with collision intensity. We also found that a simple forest variable performed almost as well as the complex connectivity model. WVC hotspots did not differ from adjacent roads or regional roads in terms of connectivity, except that traffic volume was higher at hotspots. Our findings suggest that the relationship between habitat connectivity and WVCs depends on broader landscape context and likely exhibits nonlinearity. Our work also demonstrates that some connectivity models are better predictors of WVCs than others, emphasizing the role of species-specific habitat connectivity assessments. These results can inform WVC mitigation planning and enhance understanding of habitat connectivity's role in broader landscapes.

Wildlife-vehicle collisions (WVCs) impose serious and increasing environmental, economic, and societal costs worldwide. Examining temporal and spatial patterns in WVCs is a critical piece of the wildlife connectivity puzzle that can help reduce the frequency and severity of collisions for humans and wildlife. We analyzed WVC records in New Hampshire (USA) between 2002 and 2019 to visualize spatiotemporal patterns, evaluate statistical predictors of WVCs, and to identify priority areas for mitigation efforts. More than 27,000 WVCs were reported between 2002 and 2019 throughout the entire state, averaging approximately 1500 WVCs per year. WVCs occurred on roads of all functional classes; notably, 33

This is the story of the largest movable bridge replacement in NH’s history, a compelling account of new ideas to match and exceed an historically significant 90-year old lift bridge in design, style, and innovation with today’s unique high-performance features designed to last a century. These bridge features include a first-in-the-world structural design, and the first use of thermal spray zinc coating in New Hampshire and Maine, whose pewter-colored finish blends with the naval and marine river setting, and whose success has encouraged the growth of metallizing in shops and on bridges in New England over the past decade.

What is the best paint to use when long-used lead-containing paint is no longer permitted, when VOC limits are lowered, when manufacturers are developing new systems, and coating performance is expected to last decades? Such were the questions in 1992 when NEPCOAT, the New England (now Northeast) Protective Coating Committee, was founded to evaluate and qualify the best performing paint systems. For thirty years NEPCOAT has persevered in this purpose and biannually publishes a Qualified Products List. Their testing specification grew from regional to national significance when AASHTO adopted and expanded it to administer lab testing by the National Testing Product Evaluation Program (NTPEP). NTPEP makes test result data available through DataMine. Today many states use the NEPCOAT QPL and NTPEP DataMine in the selection of bridge paint systems. This paper unfolds the story of NEPCOAT.