Experiential Mapping and Localization
An accurate mapping and localization system acts as a memory module which greatly simplifies the tasks of perception and planning by putting ego car in a context. While classical SLAM (Simultaneous Localization and Mapping) solves this problem to some extent, it leaves many unrealistic assumptions for autonomous driving, e.g. static environment, similar visual appearances between sessions, etc. We aim to address these challenges by the “experiential mapping” concept, where we draw intuition from human driving experience. The result is a more robust mapping and localization system which evolves over time and focuses on important information.