Among the main sub-areas covering the cooperative control problem of Unmanned Aerial Vehicles (UAVs), formation flight has attracted great interest and has been widely investigated. In this context, this work treats the problem of position formation flight control of a group of three Multirotor Aerial Vehicles (MAVs) under obstacle and collision avoidance constraints. In order to solve the problem, a decentralized architecture with Model Predictive Controllers (MPC) for each vehicle includes a set of convex constraints on the vehicles's position to prevent collisions with other vehicles and obstacles of different shapes. The resulting decentralized scheme controls the formation based on a virtual structure approach where the computers of the architecture exchange position data through diagrams in Simulink. The formation control strategy takes other vehicles predicted positions into account for collision avoidance purpose. The performance of the method is assessed through numerical simulations considering that the vehicles are subject to disturbance forces and the results show the effectiveness and the ability of the formation control architecture to handle the obstacle and collision avoidance constraints. For experimental validation of the proposed formation control, a estimation system based on cameras was developed to determine the position and velocity of the vehicles. In this experiment, are used two Parrot AR.Drone 2 platforms as elements of the fleet and a static Kinect sensor for Windows, fixed on the ceiling of the indoor arena and pointing downwards, composing a low-cost formation control experiment.