One key transportation problem is that of pickup and delivery; in other words, how to deliver goods from one location to another. Some of the aspects to consider are the acceptance/rejection of orders and the assignment of freight to vehicles. We tackle these problems considering the context of logistics from a European country. To do so, we built an environment from scratch, with heterogeneous trucks and freight that are randomly generated. To solve the environment, we utilized Reinforcement Learning (RL) approaches from Python’s Imitation Library and compared them amongst themselves and with simple heuristics. The criteria we used to judge the approaches were: distance travelled, distance travelled with no freight, and grams of CO2 emitted. As for the heuristics, heuristic 1 simply matches with the first option that is possible, while heuristic 2 goes through all the possibilities for matching a freight with every possible truck and picks the “best” one depending on the aforementioned criteria. We find that although the Imitation Learning approaches have issues in balancing the distance travelled with the CO2 emissions, they do well for the distance metrics, performing better than heuristic 2 on that front. We also conclude that approaches using real human input might not yet be ready for real-world deployment.