How can we leverage predictive models to overcome error in visual odometry? This code is the start of implements a baseline MDN with an LSTM to predict vehicle position with sensory input. We collect driving examples from a human controlling a remote and sim car and use these traces to improve state estimation. More to come...
tedrepo / planning_vo Goto Github PK
View Code? Open in Web Editor NEWThis project forked from johannah/planning_vo
Learned state estimation with MDN+RNN and Visual Odometry. Work in progress.