The Multiple-Depot Vehicle Scheduling Problem (MDVSP) is very important in the planning process of transport systems. It consists in assigning a set of trips to a set of vehicles in order to minimize a certain total cost. We introduce three fast and reliable heuristics for MDVSP based on a classical integer linear programming formulation and on graph theoretic methods of fixing the infeasible subtours gathered from an integer solution. Extensive experimentations using a large set of benchmark instances show that our heuristics are faster and give good or even better results compared to other existing heuristics.