Switching Q-learning in partially observable Markovian environments | IEEE Conference Publication | IEEE Xplore