幼苗期宿根蔗田机器人自主循垄导航方法

Autonomous Robotic Row-following Navigation Method for Ratoon-emerged Sugarcane Seedling Fields

摘要:

研究幼苗期宿根蔗田机器人自主循垄导航方法，对于推进宿根蔗田幼苗期管护智慧化升级至关重要。本研究在前期甘蔗幼苗期垄向导航线提取的研究基础上，提出了基于时空记忆补偿的循垄导航路径规划策略，搭建了轻简机器人系统，建立了幼苗期宿根蔗田机器人自主循垄导航方法。针对缺苗断垄挑战性场景，在 0.2m/s 的优选行进速度下，当设定路径拟合区间为有效视场的 2/3 时，循垄导航的精度较高，机器人横向偏差的均方根误差为 4.41cm, 表明机器人系统对图像噪声和蔗田缺苗断垄干扰具有较好的鲁棒性。机器人系统的平均帧率达 17.3f/s, 其中逻辑决策与控制平均耗时仅 4.7ms, 表明本研究视觉模型、导航线提取算法和边缘端构建的自主 "能力栈" 能够满足机器人在非结构性蔗田环境下快速自主循垄的时效需求。

Abstract:

It is very important to research the autonomous robotic row-following navigation method of ratoon-emerged sugarcane field robots at the seedling stage to promote the intelligent upgrading of management and protection in the seedling stage of ratoon-emerged sugarcane fields. On the basis of the research on the extraction of row-following navigation lines at the early stage of sugarcane seedlings, a row-following navigation path planning strategy was proposed based on spatio-temporal memory compensation, a light robot system was built, and an autonomous row-following navigation method for perennial sugarcane field robots at the seedling stage was established. At the optimal travel speed of 0. 2 m/s, when the set path fitting interval was 2/3 of the effective field of view, the accuracy of rowfollowing navigation was high, and the root mean square error was 4. 41 cm, indicating that the robot system exhibited robustness in image noise and seedling absence in sugarcane fields. The average frame rate of the robot system was 17.3 f/s, and the average time of logical decision-making and control was only 4. 7 ms, indicating that the vision model, navigation line extraction algorithm and autonomous "capability stack" constructed at the edge can meet the timeliness requirements of robots to quickly and autonomously row following in an unstructured sugarcane field environment.