Dealing with multiple experts and non-stationarity in inverse reinforcement learning: an application to real-life problems.