The computation of alternative paths for point-to-point shortest paths on time-dependent road networks has numerous practical applications. Despite its importance, there has been a lack of research in the literature addressing alternative paths in time-dependent road networks. In this paper, we present an innovative algorithm designed to generate high-quality alternative paths in a time-dependent context. Our approach leverages an existing time-dependent bidirectional A* algorithm. We first introduce an efficient method for gathering candidate alternative paths by identifying intersections between forward and backward searches. We then present a filtering approach for these candidate paths to ensure that only the highest quality alternatives are returned to the user. Simulation results confirm that our approach achieves good latency performance, returns optimal time-dependent paths, and provides high-quality alternative paths.

Modern direction services must evolve to meet the growing complexity of user queries, which increasingly resemble natural language and include nuanced constraints and preferences. This paper introduces a dynamic, query-aware route enrichment framework designed to enhance routing services by integrating real-time contextual data—such as weather, events, and POIs. The system comprises five key components: query intent understanding, hint point selection along the route, data sourcing, language model-based response generation, and caching for performance optimization. A hybrid approach combining lightweight language models and a rule-based system is used to interpret user intent, while adaptive hint logic minimizes redundant API calls. The enriched route responses can be consumed directly by user interfaces or serve as grounding data for LLM-based assistants. A demo application illustrates the system's effectiveness, showing reduced latency and high response quality. This work demonstrates the feasibility of real-time, intelligent route enrichment and lays the groundwork for future enhancements using AI agents and parallelized architectures.

Searching for directions is one of the most used features of map applications. This paper shares our vision on how Direction Services will change, with LLM-based chat assistants rapidly becoming an integral part of the underlying path search mechanism. We anticipate an influx of more complex, conversational route planning sessions, where users colloquially describe route-related preferences as if they were talking to their personal chauffeur. We envision future systems able to support asks like "avoid the East River tunnel", "take the bridge", or "find me a scenic route around the lake, oh and by the way, I'm driving the EV today". At present, popular map search engines fail in even simple, yet very natural preferences, such as 'take me from A to B via road C'. The reason is mainly twofold, inadequate query understanding and lack of mechanisms in routing to satisfy this type of preferences. The here proposed solution is a novel treatment of routing, one which casts it into an end-to-end 2-layer relevance framework. The framework is capable of performing query understanding for route queries with complex preferences and intents. It treats routes as richly annotated documents and the routing engine, in addition to performing optimization, acts (1) as a retriever of route documents that match the user intent and (2) as a ranker that ranks route candidates not just by a simple time-distance cost model, but by inferring the importance of many variables, some derived from explicitly stated preferences and others identified as relevant through data-driven methodology.

People often ask questions for which the answer contains multiple locations or locations with additional context. The questions can be very natural and easy to understand by other people, yet if formulated as map queries, today's map search engines struggle to understand them. Here we look into three categories of such map queries: 1) queries with multiple explicitly stated locations, e.g. 'Show me directions from A to B through C, D, and E'; 2) queries where the locations need to be inferred, e.g. 'Show me on the map all locations which James Bond visited in Casino Royale'; 3) queries of locations where we request additional geographical, historical or other context, e.g. 'Show me a map of wildlife in Australia'. We build a prototype system, called Map GPT Playground, and demonstrate with it how such queries can be seamlessly answered by combining the power of Large Language Models (LLM) with foundational maps services, such as geocoding, routing, etc. We describe the architecture of the system and reason over the abstractions needed for the system to be able to generalize across complex query intents and invoke suitable chains of services to fulfil them. Lastly, we demonstrate that in resolving complex location search queries, novel considerations emerge without prior analog, namely the set of returned locations needs to be spatially consistent and often to satisfy some inferred from the query temporal order.

Map search engines compute the estimated time of arrival (ETA) from location A to location B by first performing local routing-engine optimization over a network of road segments. Once the optimal route candidates are identified their ETA is reevaluated with global post-routing ETA (PostETA) models capable of correcting multiple accumulated local biases. Sequence models have emerged as the state of the art post-routing ETA predictors, however, they are usually applied as regressors fitting a single ETA value. Here we demonstrate that a route can have very different travel times for different drivers even when measured at approximately the same starting time. Fitting a distribution then, instead of a single value, and returning to users both an expectation over ETA together with a confidence range is more accurate and informative. We propose a novel PostETA system including a set of sequence-to-sequence attention models capable of fitting the route ETA distribution. On a data set of over a hundred thousand user trips we demonstrate that the system achieves accuracy comparable to that of regression models, providing in addition an accurate estimate for the variance of the ETA prediction.

Recent machine learning models find a widespread use whether in respect of data mining and forecasting or in the classification domain. However, real-world situations comprise complex estimation tasks that carry a certain semantic load and bring a certain degree of fuzziness with them. This is a fuzziness which humans, due to their common sense knowledge and their personal experience, can easily understand by linking the underlying concepts together, while machines may from scratch not. A vast amount of both training data and time are necessary in order for a computational model to be capable of learning such kind of relations and adapting to new situations. In this work, we show that letting explicit semantic knowledge flow into a predictive model leads to an improved performance with regard to training time, accuracy and robustness. In particular, we propose adding an auxiliary semantic layer to the model, whose role is to provide it with information about the semantic interrelation of the treated classes creating in this way shortcuts and saving valuable training time while improving its quality at the same time. We explore several versions of our approach and we illustrate their functionality in a semantic location prediction scenario using 2 different real-world datasets.

Proactive mobile applications and services have the advantage of providing their users with timely and customized solutions improving in this way the human-machine interaction. For this reason, Location Based Services (LBS) rely increasingly on predictive models that estimate how likely it is for a user to visit a certain location. Recently, Artificial Neural Networks, and especially recurrent architectures such as the LSTMs, have shown a particularly good performance in this field. In this work, we extend a LSTM network by applying Sequence to Sequence (Seq2Seq) learning on human semantic trajectories. In particular, we explore whether and to what extent Attention-based Seq2Seq learning in combination with neural networks can contribute to improving the accuracy in a location prediction scenario. We compare the performance of our framework with the performance of a standard LSTM, a semantic trajectory tree-based approach and a probabilistic graph of first and higher order on two different real-world datasets. It can be shown that Sequence to Sequence learning may well be used to model semantic trajectories and predict future human movement patterns.