Abstract:

Based on a segmented uniform acceleration model, a global velocity planning method was proposed to achieve intelligent connected vehicle navigation in complex and unstructured road scenarios. By analyzing the vehicle’s dynamic characteristics with driving safety and smoothness as the principles, the critical speeds for rollover and sideslip were calculated as the maximum traveling speeds for each path point on unstructured roads. The global velocity planning problem was formulated using a segmented uniform acceleration model, taking into account efficiency, smoothness and energy consumption through a comprehensive loss function. The model considered the continuous variation of slope curvature on unstructured roads and designs variable bounds for vehicle velocity, acceleration, and jerk to constrain the decision variables. By integrating the regenerative braking function of heavy-duty electric vehicles, a specific velocity planning model for electric vehicles in unstructured road scenarios was proposed, and the method was validated through simulations. The results show that the acceleration range of the ego vehicle is stable within -1.0―1.0 m/s², and the jerk range is stable at -0.5―0.8 m/s³. Compared with the speed planning based on dynamic programming, the proposed speed planning algorithm not only ensures the stability of the vehicle but also reduces the vehicle control inputs. The proposed method has been applied to the autonomous trucks which travel smoothly, the maximum jerk of the truck does not exceed 0.45 m/s³, which shows the stability of the truck.

Key words: unstructured road, speed planning, autonomous driving, motion planning