This thesis focuses on novel approaches for dealing with last-mile operations faced by logistics service providers in urban contexts. The last-mile refers to the last link in the transport chain followed by a parcel to fulfill consumers’ requests for goods, from the shelf of the last distribution center to the hands of the buyer. We investigate two recent innovations and the potential cost-benefits of introducing such models into transportation logistics for last-mile operations. More specifically, we first consider a crowd-sourced solution – where drivers are not employed by a carrier but occasionally offer their services through on-line platforms and are contracted as required by the carrier – for the fulfillment of transportation requests and evaluate the benefits of introducing transfers to support driver activities. We frame the problem as an extension of a pickup and delivery problem with transfers and propose a heuristic optimization method to solve it. The second novel model we consider is what has been defined as roaming delivery systems, in which the service provider has access to private cars’ storage compartments, and can service customers using the trunk of their cars. Supported by automotive and communication technologies, the model has the potential to make e-commerce operations more convenient, mitigating failed deliveries at home. We introduce a stochastic version of the Vehicle Routing Problem with Roaming Delivery Locations (VRPRDL) and propose a two-stage stochastic model using the possibility of servicing customers at different locations as a recourse action. Finally, we introduce a dynamic variant of the VRPRDL in which customers announce in real-time the locations where their cars are or will be parked and the service provider decides whether visiting customers at home or at their roaming locations.