Emanuele Manni

Real-time navigation systems are increasingly used to provide optimal driving routes together with accurate travel time predictions that reflect dynamic urban traffic conditions. Recent advances have focused on integrating structured traffic data from traditional APIs with unstructured, context-rich information extracted via semantic crawling of news websites and social media platforms. This survey reviews state-of-the-art approaches that combine these heterogeneous data sources to improve route planning and travel time estimation, with special attention to the challenges posed by incident detection, event extraction, and multimodal data fusion. We discuss core methodologies including natural language processing techniques for event recognition, machine learning models for traffic prediction, and graph-based routing algorithms, highlighting their advantages and limitations. Finally, we outline open research directions for building context-aware navigation systems able to adapt to real urban mobility conditions.

This paper deals with the park‐and‐multi‐loop routing problem, in the context of last‐mile logistics, in which a fleet of traditional vehicles equipped with several autonomous delivery robots leaves from a depot to service a set of customer requests. Each delivery route involves a main tour of a traditional vehicle (subject to time‐dependent travel times), which may stop at one or more parking locations, from which one or more walking sub‐tours are carried out by the driver and the delivery robots on the pedestrian network. The goal is to minimize the overall cost of both van and walking routes. An alternative objective aims to minimize the emissions. A peculiarity of the problem is that the walking distance covered by drivers and delivery robots must not exceed given thresholds. For this problem, we propose a tailored destroy‐and‐repair mechanism embedded into a neighborhood search algorithm. The destroy‐and‐repair operators are parameterized, and the most appropriate parameter setting is chosen based on some performance measures of the solution. Computational results on a set of instances of varying size, derived from real data of two major Italian cities, certify the quality of the approach, compared to classic destroy‐and‐repair operators. Moreover, we also show the benefits of the proposed delivery scheme in terms of emissions when compared to a traditional delivery paradigm.

The growing concern over emissions from last‐mile logistics is motivating the development of innovative delivery models aimed at reducing environmental impact and traffic congestion. To this end, we propose a last‐mile delivery scheme based on both public drop‐off boxes and autonomous delivery robots. Unlike current practices, we assume that parcels are primarily delivered by a fleet of traditional vehicles to a number of drop‐off boxes owned by the municipalities and shared by several logistics providers. Then, citizens must autonomously reach the drop‐off point to retrieve their parcels. An exception is made for a small group of customers (like, for instance, elderly or disabled people) whose parcels are loaded on autonomous delivery robots and home‐delivered, starting from the drop‐off points. To tackle this problem, we devise tailored destroy‐and‐repair operators to be embedded into a neighborhood‐search framework. We then conduct an extensive computational study on a set of realistic instances based on the urban area of Rome (Italy), aimed at assessing the benefits of the proposed last‐mile delivery infrastructure, compared to traditional distribution methods. Additionally, we evaluate the environmental and economic advantages of using autonomous delivery robots and drop‐off points.

Since the 1950s, optimization-based Decision Support Systems have emerged as one of the most impactful applications of Operations Research, replacing human-based planning across a wide range of industries. These systems have greatly improved decision-making in terms of efficiency, accuracy, and scalability. However, they also come with a significant limitation: a lack of flexibility and adaptability. Typically, the algorithms used in these systems are predefined during the design phase, following a thorough - and often time-consuming - requirements analysis. They are then hard-coded into the system, making them well-suited for specific, well-understood scenarios. However, this rigidity makes it difficult for such systems to adapt to unforeseen changes or evolving operational needs. As a result, expert intervention, re-analysis, or even a complete system redesign may be required - often at considerable time and cost. This paper explores how Large Language Models (LLMs) can be integrated with traditional optimization algorithms to address this key limitation - lack of flexibility - while preserving their core strengths: speed and solution quality. The core idea is to leverage LLMs to interpret natural language instructions, reconfigure algorithmic components, support preference-based decision-making, and explain the rationale behind their choices. Computational results on a "rich" Vehicle Routing Problem (VRP) setting - a class of VRPs that incorporate multiple real-world constraints and complexities commonly encountered in last-mile distribution - demonstrate the potential of this hybrid approach.