A systematic review of machine learning models for road and last-mile freight optimization under cost and sustainability objectives in multimodal transportation systems

Road-based freight and last-mile delivery are foundational modes of any multimodal transportation system and contribute disproportionately to global energy demand and greenhouse-gas emissions. Machine learning (ML) has emerged as a methodological bridge between operational data and the joint cost / sustainability objectives the sector must improve. This paper reports a PRISMA 2020–compliant systematic literature review of ML models applied to road-freight and last-mile transportation optimisation with explicit cost or sustainability outcomes, published between 2015 and 2025. The 2015 start point was chosen because a preliminary scoping search returned negligible relevant output before 2015; within the review window, annual output rises from a single eligible study in 2015 to 169 in 2025 (Sect. 3.1). Two databases (Scopus and Dimensions) yielded 7,593 records, reduced to 4,531 after deduplication and auto-filtering and then to 361 included studies after a four-criterion title/abstract screening. We performed bibliometric description, unsupervised KMeans thematic clustering on TF-IDF of title+abstract+keywords, three co-occurrence networks (author keywords, co-authorship, country collaboration), and a methodology × application cross-tabulation. Four themes structure the literature: (i) emissions and fuel-consumption prediction (n = 103); (ii) deep reinforcement learning (DRL) for vehicle routing (n = 70); (iii) last-mile and urban-logistics optimisation (n = 130); and (iv) electric-vehicle energy management and eco-driving (n = 58). The methodological centre of gravity has shifted to deep learning (51% of the corpus) and DRL (23%); DRL grew from absent in 2021 to 47 papers in 2025. We identify six research gaps and propose a five-point research agenda with explicit implications for the integration of road and last-mile operations into multimodal transportation systems.