The work aims to propose a new green delivery mode using mobile parcel lockers and onboard robots for collaboration to improve the 'last mile' delivery efficiency. First, considering customer preferences, time windows, the vehicle capacity and the coordination between mobile parcel lockers and delivery robots, a route optimization model was built with the objective of carbon emissions minimization. Second, an adaptive large neighborhood search algorithm was developed to solve the corresponding routing problem. Finally, numerical experiment analyses were conducted using instances generated from a real case. The proposed algorithm obtained the same or better solution as the Gurobi solver under the time limit of 1 h on fifteen instances. Three out of ten instances obtained different solutions when taking cost minimization as objective. Among them, solutions under carbon emissions minimization were with about 0.3% higher cost and 2.41% lower carbon emissions. In terms of carbon emissions and delivery costs, the collaborative delivery using mobile parcel lockers and onboard robots could achieve a reduction of 53.29% and 33.38% when compared with truck-only system, a reduction of 41.85% and 62.01% when compared with the fixed parcel lockers system, and a reduction of 37.80% and 32.76% when compared with the vehicle mobile parking for the self-pickup system. Benefits of the proposed delivery system are validated. The proposed route optimization model and the designed solution algorithm can provide a valuable reference and decision support for building a low-carbon sustainable urban delivery system.