Decomposing the Prediction Problem; Autonomous Navigation by neoRL Agents