Monocular based lane-change on scaled-down autonomous vehicles

Lane-change is considered to be one of the toughest challenges in the field of autonomous vehicles. Any vehicles that are driving around the host vehicle may effect this operation. To ascertain the safety during the lane-change process, a comprehensive understanding of surrounding environments as well as a real-time decision is needed. Many researchers have proposed several approaches based on kinematics to perform this task. However, human drivers does not concerned too much about the kinematic parameters of vehicles, but try to control vehicles mainly based on the scenes from vision and experiences accumulated from long-time driving. From this perspective, we propose a more human-like lane-change system, whose sensors mainly based on five monocular cameras. Information from these sensors are fused into MOR (My Own Range) which means the drivable area for the vehicles. Finally, two fuzzy controllers are used to imitate expert behaviors and responses upon vehicle control. To verify the efficiency of the proposed lane-change system, a test platform, which includes ten scaled-down autonomous vehicles and a scaled model highway, is built. With this platform, tons of ideas or tests, especially of some dangerous tests, can be examined and implemented with a low cost later.