考虑经验驾驶行为的入弯实时类人速度规划方法

Real-time human-like speed planning method for curve entry considering experienced driving behaviors

为提高自动驾驶汽车的安全性、舒适性与通行效率,提出了一种考虑到滑行及安全速度的类人入弯速度规划策略。该策略基于混沌优化理论与实车弯道行驶速度数据,通过将车辆的入弯速度规划问题构建为多目标优化问题,建立了舒适模式与效率模式。通过定义奇点速度,简化了非线性高阶约束条件。结果表明:在该策略规划中,在不同弯道场景下,横向加速度和纵向加速度均在摩擦圆约束内,可保证车辆的行驶安全。相比于未考虑滑行的方法,在舒适模式下,该策略产生的纵向加速度减小9.76%,通行效率提高61.73%;在效率模式中,纵向加速度为加速度阈值,满足加速度约束,通行效率提高88%。因而,无论舒适模式还是效率模式,该方法均可兼顾舒适性和通行效率。

A human-like safe-speed-planning method was proposed for vehicle curve entry considering coasting and safe-speed based on the chaos theory and the real curve driving velocity data to improve the safety,the comfort and the travel efficiency of autonomous vehicles.A comfort mode and an efficiency mode were constructed through formulating the curve-entry speed-planning problem of autonomous vehicles as the multi-objective optimization problems.A singularity velocity was defined to simplify the constraint condition with high-order and non-linear characteristics and to improve its computation efficiency.The results show that both the lateral accelerations and the longitudinal accelerations satisfied the friction circle constraint for the proposed strategy,with guaranteeing the driving safety in different curve scenarios.Compared with the method regardless of coasting,the maximum longitudinal acceleration, which are generated by the proposed strategy, was reduced by 9.76% in the comfort mode with the travel efficiency being improved by 61.73%. In the efficiency mode, the longitudinal accelerations values are the acceleration threshold, which satisfy the acceleration constraint with an increase of 88% in traffic efficiency. Therefore, both the comfort mode and the efficiency mode achieve a balance between comfort and travel efficiency.