An unmanned driving system based on lane-level path planning

Abstract

Automated driving systems promise low cost and low human consumption. If it is used in mine, canyons and other environments, it will have huge economic benefits. However, in such environments as mines and urban canyons, there is a problem that satellite signals are blocked, leading to the failure of positioning. To solve this problem, we integrate lidar, inertial measurement unit and Real-Time Kinematic Global Position System to achieve high-precision positioning in urban canyon and open environment. Besides, there are many curves on the roads in urban parks, which adds great difficulty to unmanned driving, so we construct a lane-level high-precision environmental map, which realizes path planning based on lane and stable driving of unmanned vehicles. Furthermore, we orderly integrate perceiving, mapping and positioning, path planning and motion control modules to form a lightweight unmanned driving system, which perceive the environment by lidar, inertial measurement unit and Real-Time Kinematic Global Position System, use lightweight SC-LEGO-LOAM to build environment map, use normal distribution transformation to achieve rapid vehicle positioning, and use lane-level high-precision map to achieve global static path planning, use lattice algorithm to realize smooth and stable local path planning, then transmit it to the vehicle site. After real vehicle testing, the vehicle can be driven stably in the complex environment of the park. This automated driving system can be applied in mines and urban parks and can realize unmanned transportation. It has huge economic benefits. The lane-level high-precision map we have built is the development direction of the future driverless electronic map.