Collision avoidance requires tradeoffs in planning time horizons. Depending on the planner, safety cannot always be guaranteed in uncertain environments given map updates. To mitigate situations where the planner leads the vehicle into a state of collision or the vehicle reaches a point where no trajectories are feasible, we propose a continuous collision checking algorithm. The imminent collision checking system continuously monitors vehicle safety, and plans a safe trajectory that leads the vehicle to a stop within the observed map. We test our proposed pipeline alongside a teleoperated navigation in real-life experiments, and in simulated random-forest and warehouse environments where we show that with our method, we are able to mitigate collisions with a success rate of at least 90\%.
翻译:避免碰撞需要在规划时间范围内进行权衡。 取决于规划者, 在地图更新的不确定环境中, 安全并不总是能得到保证。 为了减轻计划者引导车辆进入碰撞状态或车辆到达一个没有轨道可行的地点的情况, 我们提议一个连续的碰撞检查算法。 即将到来的碰撞检查系统持续监测车辆安全, 并计划一个安全轨道, 引导车辆在观察到的地图内停留。 我们测试了我们提议的管道, 并同时进行实际实验中的远程操作导航, 以及在模拟随机森林和仓库环境中, 我们用我们的方法显示, 我们能以至少90个成功率来减少碰撞。