This paper presents a Distributed Model Predictive Control approach for UAVs trajectory optimization with the state-dependent collision avoidance criterion. While classical trajectory tracking optimization criterion ensures accuracy of arrival points for the multi-agent system, risk of collision is evaluated at every optimization step, and collision avoidance is included or neglected in the cost function accordingly. Proposed solution is finally evaluated in simulation tests with two scenarios, intersecting and parallel paths of two UAVs. It is, therefore, applicable for different planned path configurations.