Deepening method for LeanMAP content based on a virtual trajectory by lateral transcription

Abstract

In this study, we aim to develop a new upgrading method of digital map content for automated driving on nationwide public roads. In general, a dense waypoint map based on a digital map is necessary to realize precise motion controls of intelligent automobiles. However, current preparation methods for digital maps have two problems for practical use on public roads in wide areas: the size of the map data for onboard storage and the monetary and human resources for manual mapping processes. To solve these problems, in our previous study, we proposed a new digital map called “LeanMAP,” which consists of elementary information about the road, and also proposed an upgrading method of its contents from the level of car navigation map to the level of quasi-precise map. Furthermore, in this study, we propose another upgrading method based on a virtual trajectory to reduce the number of resources necessary for the mapping process to as low as possible. First, we propose a lateral transcription method that generates a virtual trajectory of an adjacent lane. Although three or more driving data items are necessary in order to upgrade the LeanMAP contents in the previous method, the proposed method in this study requires only one item of driving data. Then, to evaluate the accuracy of the dense waypoint map based on the elemental information upgraded by the proposed method, we conduct an initial evaluation using actual driving data on a public road. As a result, we confirm that the proposed system satisfies the target value of less than 1 % lateral error. From this result, we confirm the feasibility that our proposed system can reduce the cost and manual operations of mapping a precise digital map.

Keywords

• automobile
• intelligent vehicle
• digital map
• virtual trajectory
• map deepening