This paper presents a truck routing and scheduling problem faced by a public works company. It consists of optimizing the collection and delivery of materials between sites, using a heterogeneous fleet of vehicles. These flows of materials arise in levelling works and construction of roads networks. As the quantity of demands usually exceeds the capacity of a truck, several trucks are needed to fulfill them. As a result, demands are split into full truckloads. A set of trucks routes are needed to serve a set of demands sharing a set of resources, available at pickup or delivery sites, which can be loaders or asphalt finishers in our application cases. Thus, these routes need to be synchronized at each resource. We propose an Adaptive Large Neighborhood Search (ALNS) to solve this problem. This approach is evaluated on real instances from a public work company in France.