Research on a working station for autonomous bio-production robots

为了克服日本农业面临的困难，研究了自主行走技术，作为开发农业机器人的基础。为了利用农业机械完成全自动化作业的开发，必须实现自动装填农用物资或自动更换农具。研究了化肥灌装容器或机具等待接近物体的搜索方法和接近的控制方法，提出了利用简单的激光传感器进行三角测量的物体定位方法。经实际试验验证，所研制的定位系统在7m以内的距离内定位精度小于10 cm，所研制的进近控制基本上是开环控制，将车辆引导到预先给定条件下设计的轨迹上。首先证明了利用车辆实轴中心的线性化运动方程，可以在给定的初始条件下设计弹道。虽然以同样的方式推导出了安装在车辆上的执行机构的轨迹方法，但控制参数必须通过搜索来确定。此外，通过仿真研究验证了该方法的可行性。最后，利用实际尺寸的自主拖拉机进行了田间试验，验证了控制器的性能。结果表明，在混凝土路面测试范围内，目标点的最终位置误差小于1%。作为图像处理的结果，从噪声图像中检测边界。而图像处理的另一种应用使我们能够发现周围区域的运动物体。它可能适用于自动驾驶汽车的安全装置。

To overcome facing difficulties of Japanese agriculture, autonomous traveling technology was investigated as a base of development of robots for agricultural use. For completion of development of fully automated operations by using agricultural machines, it is requisite to realize automatic refilling of agricultural material or automatic exchange of implements by an autonomous vehicle itself. In this study we concentrated on a searching method of an object to approach, such as a refilling container of fertilizer or an implement to be mounted, and a control method to approach it.The positioning method of an object that used triangulation surveying with a simple laser sensor was developed. It was confirmed by actual experiment that the accuracy of the developed positioning system was less than 10cm in the distance less than 7m.The developed approaching control is basically an open loop control which steers the vehicle on the trajectory designed from given conditions previously. At first it was shown that we can design a trajectory from given initial conditions by using linearized motion equations of the center of the vehicle's real axle. Though the trajectory method for the implement mounted on the vehicle was derived in the same manner, the control parameters have to be determined by a searching. In addition the feasibility of the methods were checked through simulation studies. Finally the performance of the controller was examined in field experiments by using an autonomous tractor of actual size. As a result the final positional error at the target point was less than 1% in the range of tested on concrete roads.As a result of image processing, detection of boundary from a noisy image was developed. And the other application of image processing enabled us to find out moving objects in the surrounding area. It might be applicable as a safety device for autonomous vehicles.