Abstract
This work presents a deep learning framework for analyzing urban mobility by extracting knowledge from messages collected from Twitter. The framework, which is designed to handle large-scale data and adapt automatically to new contexts, comprises three main modules: data collection and system configuration, data analytics, and aggregation and visualization. The text data is pre-processed using NLP techniques to remove informal words, slang, and misspellings. A pre-trained, unsupervised word embedding model, BERT, is used to classify travel-related tweets using a unigram approach with three dictionaries of travel-related target words: small, medium, and big. Public opinion is evaluated using VADER to classify travel-related tweets according to their sentiments. The mobility of three major cities was assessed: London, Melbourne, and New York. The framework demonstrates consistently high average performance, with a Precision of 0.80 for text classification and 0.77 for sentiment analysis. The framework can aggregate sparse information from social media and provide updated information in near real-time with high spatial resolution, enabling easy identification of traffic-related events. The framework is helpful for transportation decision-makers in operational control, tactical-strategic planning, and policy evaluation. For example, it can be used to improve the management of resources during traffic congestion or emergencies.

Abstract
E-commerce growth is raising the demand for logistic activities, especially in the last-mile, which is considered the most ineffective part of the supply chain and a negative externalities source. Although various solutions aim to address these issues, selecting the best one is challenging due to multiple perspectives, conflicting criteria, trade-offs, and complex and sensitive urban contexts. This article proposes a 4-level hierarchical model based on the triple bottom line of sustainability that may assist decision-makers in selecting the most adequate last -mile solution for historic centers. The model was defined based on a systematic literature review; evaluated by interviewing a set of experts; and quantified according to an AHP-TOPSIS approach. This quantification focused on the historic center of Porto, Portugal. The experts considered all three sustainability dimensions similarly important. Air pollution was the most valued sub-criterion whereas Visual pollution was the least. 67 decision-maker profiles were defined, showing that environmentally oriented decision-makers prefer cargo bikes, while decision-makers who prioritize economic and social factors prefer parcel lockers. All last-mile solutions considered in the model yielded similar results, therefore suggesting a combined distribution strategy. Nevertheless, the use of parcel lockers is the most favorable solution for Porto's historic center.

Abstract
Globalisation, urbanisation and the recent COVID-19 pandemic has been raising the demand for logistic activities. This change is affecting the entire supply chain, especially the last-mile step. This step is considered the most expensive and ineffective part of the supply chain and a source of negative economic, environmental and social externalities. This article aims to characterise the sustainable urban last-mile logistics research field through a systematic literature review (N = 102). This wide and holistic review was organised into six thematic clusters that identified the main concepts addressed in the different areas of the last-mile research and the existence of 14 solutions, grouped into three types (vehicular, operational, and organisational solutions). The major findings are that there are no ideal last-mile solutions as their limitations should be further explored by considering the so-called triple bottom line of sustainability; the integration and combination of multiple last-mile alternative concepts; or by establishing collaboration schemes that minimise the stakeholders' conflicting interests.

Abstract
Globalization and the COVID-19 pandemic led to an increased number of consumers using e-commerce services. This trend has been raising the demand for logistic activities, especially on the last-mile. This part of the supply chain is expensive and ineffective, and a source of negative externalities such as air and noise pollution, traffic congestion and accidents. The anticipation of innovative solutions can help to mitigate these costs. In this context, this paper provides a systematic literature review of the existing literature regarding emerging solutions for last-mile parcel delivery. For guiding the development of more sustainable last-mile parcel distribution, and to provide some insights for future research, we identified and summarized the emerging concepts within this field domain. The results show that innovative solutions have been emerging at different levels: (i) definition of new crowdsourcing-based models, (ii) use of new types of vehicles, and (iii) development of optimization systems based on data collection and the combination of different technologies. Moreover, recent studies show that new strategies are being developed focusing on using consumers as active actors of delivery; non-road and autonomous vehicles are promising concepts in last-mile operations; and different logistic operations, such as vehicle routing, are being optimized with data analytics, cloud technology and mobile apps.

Abstract
The Vehicle Routing Problem with Divisible Deliveries and Pickups (VRPDDP) is under-explored in literature, yet it has a wide application in practice in a reverse logistics context, where the collection returnable items must also be ensured along with the traditional delivery of products to customers. problem considers that each customer has both delivery and pickup demands and may be visited twice in the same or different routes (i.e., splitting customers' visits). In several reverse logistics problems, capacity restrictions are required to either allow the movement of the driver inside the vehicle to arrange the loads or to avoid cross-contamination between delivery and pickup loads. In this work, explore the economic and the environmental impacts of the VRPDDP, with and without restrictions the free capacity, and compare it with the traditional Vehicle Routing Problem with Simultaneous Deliveries and Pickups (VRPSDP), on savings achieved by splitting customers visits. An exact method, solved through Gurobi, and an ALNS metaheuristic are coded in Python and used to test well-known and newly generated instances. A multi-objective approach based on the augmented e-constraint method is applied to obtain and compare solutions minimizing costs and CO2 emissions. The results demonstrate that splitting customer visits reduces the CO2 emissions for load-constrained distribution problems. Moreover, savings percentage of the VRPDDP when compared to the VRPSDP is higher for instances with a random network than when a clustered network of customers is considered.

Abstract