Self-driving vehicles rely on sensory input to monitor their surroundings and continuously adapt to the most likely future road course. Predictive trajectory planning is based on snapshots of the (uncertain) road course as a key input. Under noisy perception data, estimates of the road course can vary significantly, leading to indecisive and erratic steering behavior. To overcome this issue, this paper introduces a predictive trajectory planning algorithm with a novel objective function: instead of targeting a single reference trajectory based on the most likely road course, tracking a series of target reference sets, called a target funnel, is considered. The proposed planning algorithm integrates probabilistic information about the road course, and thus implicitly considers regular updates to road perception. Our solution is assessed in a case study using real driving data collected from a prototype vehicle. The results demonstrate that the algorithm maintains tracking accuracy and substantially reduces undesirable steering commands in the presence of noisy road perception, achieving a 56% reduction in input costs compared to a certainty equivalent formulation.

翻译：自动驾驶车辆依赖传感器输入来监测周围环境，并持续适应最可能的未来道路走向。预测性轨迹规划以（不确定的）道路走向的快照作为关键输入。在感知数据存在噪声的情况下，对道路走向的估计可能产生显著波动，导致决策犹豫和转向行为不稳定。为解决这一问题，本文提出一种采用新型目标函数的预测性轨迹规划算法：该算法不再基于最可能道路走向追踪单一参考轨迹，而是考虑追踪一系列称为“目标漏斗”的参考目标集。所提出的规划算法整合了道路走向的概率信息，从而隐含地考虑了道路感知的定期更新。我们通过使用原型车辆采集的真实驾驶数据进行案例研究来评估该方案。结果表明，在存在噪声道路感知的情况下，该算法在保持跟踪精度的同时，显著减少了不良转向指令，与确定性等价形式相比实现了56%的输入成本降低。