State and action space construction using vision information | IEEE Conference Publication | IEEE Xplore