Decision Making and Local Trajectory Planning for Autonomous Driving in Off-road Environment

Different from urban environment with high precision map, off-road environment lacks rich priori road structure information and precise location information, which is a huge challenge for autonomous driving. To solve these problems, this paper proposes a real-time decision-making and local trajectory planning method based on a predefined global reference line. According to the reference line, a cluster of parallel offset curves with corresponding obstacle and deviation costs are generated. Then, an appropriate parallel line with minimum cost is selected as the preliminary decision, and the target position with a matched speed on this line is choosed for local planning. Finally, the lateral and longitudinal trajectories planning is carried out with the initial constraint and target constraint of the quintic polynomials. The proposed method is verified in a 3.6 km bumpy off-road environment, including straight road, curves, continuous U-turn and other driving conditions. The maximum autonomous driving speed of an small-medium-sized platform can reach 30 km/h under the condition of stable and smooth driving.