Development of Robot Tractor

The robot tractor is mainly made up of a robot vehicle using commercially available tractor, so that respective parts be remodeled for automatic control, a navigation system called "XNAV" that detects and outputs robot positioning information with help of an auto-tracking type surveying device, and a controller loaded with an operation software which executes path planning and controls the robot vehicle. The robot has almost the same ability as conventional manned tractors and is capable of unmanned rotary tilling on a rectangular field at the same performance and precision level. To improve the adaptability and the utility of robot work, we developed a software that enables a path operation different from the one for conventional rotary tilling, seeding, and soil-puddling. The field test then verified the performance and the effect of the newly developed software. We proposed a double-vehicle work method by which robot operation and manned operation are performed simultaneously. With this method, an operator was proved to be able to perform a field work at an efficiency of 1.8 times as that of usual work with a conventional tractor. The result confirmed that this simultaneous method enables an effective and safe utilization of the robot. The robot performance was confirmed to be enough for its practical use from the viewpoint of labor saving, handling easiness, reliability, and work safety. Various farm works are done by the robot with work efficiency and accuracy almost at the same level as those of usual manned operation. The details of development, results of performance tests and subjects for each component are summarized below. 1. Navigation System and Robot Vehicle. For the XNAV navigation system used in this research, an auto-tracking type surveying device "AP-L1" (Topcon Corporation) is set outside the field, and it automatically tracks the light reflector placed above the center of the rear axle of the tractor, and detects the tractor positioning information by around 0.5-second time cycle. We confirmed that this system enabled measuring the tractor position at a distance of 500m with an error of 5cm or less. The tractor heading information is measured and obtained by a geomagnetic direction sensor ("TMS") mounted on the tractor. Errors due to the tractor inclination are compensated for the TMS output and measured position information. For the robot vehicle system, a commercialized tractor (Kubota GL321) is used by modifying the mechanisms such as steering mechanism, shuttle transmission, throttle and implement elevation to be controlled automatically. An automatic measurement system of the outputs of existing sensors of the base tractor and newly installed sensors was configured, so that the controller could digitally measure the conditions of operating mechanisms and each part of the vehicle. A factory computer (NEC FC9821Ka) is used for the robot main controller, and a custom made controller by Kubota Corporation is used for automatic control and measurement of each part. A safety mechanism is equipped to stop operation in case of contact to an obstacle or detection of any abnormality. 2. Robot Work and Operation Software. We designated a "basic operation method" and called a computer program to perform that operation "basic operation software". This basic operation method is based on a rotary tilling work by use of respective implement. The robot work for the basic operation method was designed to perform returning operations in the central area, excluding the peripheral area, and then the robot practiced turning around operation in the peripheral area, including the headland, in an almost horizontal rectangular field, which is similar to the case in customary rotary tilling. Information on the field lot to perform robot work is obtained through "teaching" step in which the robot is manually traveled along the operation path around the outermost periphery of the work filed lot. The operation software is composed of the "task planning section", which performs teaching and path planning, and the "traveling and operation control section", which performs traveling and operation along the generated path. The traveling and operation control section is composed of four modules: the "off-work transfer 1 module" for moving from the entrance of the field to the work starting position, the "round-trip operation module" for processing to the central area of the field lot, the "off-work transfer 2 module" for moving from the end position of round-trip operation to the initial position for turning around operation, and the "circular operation module" for turning around operation in the headland. These modules are composed of common routines for "straight forward control", "straight backward control", "180 deg turning control", "90 deg turning control", and "sideways movement control". 3. Improvement of Robot Work -Expansion of Adaptability and Reliability. To expand the adaptability of robot work, we improved the basic operation method and proposed "diagonal operation method" in which the robot returns straight along a diagonal of the field lot and "round operation method" in which the robot runs straight forward all over the field along the four sides of field lot. We modified the basic operation software, in which work is performed as in customary rotary tilling. In doing so, we secured that the method to set the work overlapping width and the order of paths were changed and selected according to work conditions such as field lot size. For diagonal operation, we developed an operation software to set an arbitrary diagonal angle on the basis of the basic operation software, by applying geometric transformation to path planning and vehicle guidance, based on the diagonal angle. For round operation, we developed an operation software so that path planning and robot work could proceed both from the outside toward the center and from the center toward the outside of the field lot. To improve the reliability and ensure the safety of robot work with XNAV, we developed programs to perform "self-diagnosis" before starting work and issue "abnormality alarming" during work, and incorporated these into the operation software. After the operator makes initial settings, the robot performs self-diagnosis in which it checks itself for ordinary acquisition of navigation data and appropriate setting and operation of each part of the robot. If it detects any abnormality, the robot requires the operator to take corrective measures. For abnormality alarming issuance, normal acquisition of navigation data during work and appropriate setting and operation of each part of the robot are checked. If any abnormality is detected, corrective measures are taken to restore the robot to the normal state based on a dialogue between the operator and the robot after the work is stopped and an alarm is issued. Operation of these functions was verified through tests on intentionally created abnormal conditions. 4. Evaluation of Robot Work Performance. We established testing methods for unmanned work and conducted evaluation tests of rotary tilling in farm fields by the robot, in order to evaluate the work performance of the robot we have discussed, integrating the navigation system, robot vehicle and operation software. We set indexes such as operator efficiency in order to evaluate the effect of unmanned work we also set indexes such as straight traveling performance and parallel traveling performance in order to evaluate the accuracy of robot work. In the evaluation test conducted in the experimental farm of IAM-BRAIN in Konosu City, Saitama Prefecture, a reference test on manned operation of the robot vehicle was also conducted. The test results indicated that it was possible to perform robot work at an efficiency almost equal to that of manned operation in a 50 * 100m field. Robot operation was superior to manned operation in terms of work accuracy such as straight traveling performance. The handling easiness of the robot was also evaluated by operators experienced in tractor work. The results indicated that the operator could perform the initial setting according to the operation manual, and could start and finish robot work without any problem. Rotary tilling tests, using the operation software developed for diagonal operation and round operation, suggested that the robot could work well without leaving untitled areas and without traveling out of the field lot. The diagonal operation and round operation method that we have proposed and developed can be performed at high efficiency and high accuracy only by the robot work. 5. Utilization of Robot Work - Application to Various Works. For an effective utilization of the robot, we proposed "simultaneous double-vehicle work method" in which an operator carries out a conventional (manned) tractor operation while executing the robot work, and we conducted work tests on it. Furthermore, we developed an operation software to perform seeding and soil-puddling work in addition to rotary tilling with the robot, and conducted working tests with it. By performing rotary tilling work with the simultaneous double-vehicle work method, it was possible for an operator to perform robot work and manned operation simultaneously. Robot work was superior to manned operation in terms of work accuracy, such as straight traveling performance. With this work method, it is possible to perform rotary tilling at a work efficiency of 1.8 times as that of manned operation using a conventional tractor; therefore, we confirmed that this method was effective and safe for utilizing robots. We prepared seeding software based on the basic operation software and conducted work tests in a farmer's field there, we confirmed that an operator performed the robot work of wheat seeding and fertilizer spreading with high accuracy including their replenishment. The evaluation confirmed that seeding with the robot could reduce labor and release farmers from hard work. Based on the basic operation software, we prepared a soil-puddling software that enables arbitrary selection and setting of repetitive work while changing travel direction. Test results confirmed that soil-puddling with robot could be carried out along an efficient route with a complete coverage. However, to improve the accuracy of soil-puddling, a function for an appropriate adjustment of work speed and tilling depth was considered to be required. 6. Subjects of Robot Work and Countermeasures. To introduce and expand the developed robot to current farm work in Japan, it is necessary to solve some remaining issues; a lower price, further enhanced safety and reliability of the robot and establishment of a method for effective utilization as well as operation of the robot at agricultural sites. The price to introduce the robot into the market from a tractor manufacturer is estimated to be about 6.5 million yen, assuming that the price of the base tractor is 2.5 million yen and the modification as well as equipment cost for converting it into a robot is 4 million yen. Employment of a GPS position detection system for the navigation system seems promising to reduce the cost of robotization. Therefore, the focus is on arrangement of the infrastructure, improvement of performance, and reduction of the utilization cost of the GPS. To ensure robot work safety, the robot must have a 100% reliable function to detect a human approaching and then to stop its work; thus, it is necessary to select human detecting sensors that have high accuracy and high reliability. The reliability of robot work can be improved by securing the detection of abnormalities and the control of robot component in detail and precisely; the operator must be able to understand such a situation from a distance and remotely execute corrective measures. It is even more desirable for the robot itself to be able to judge any abnormality and take the corrective measures. For an effective operation of the robot for farming, we studied and established a model by which an operator performs field work in a rice paddy, including transport and supply of materials. To model the robot work, it is necessary to consider methods to move the robot to the field, methods for transport and supply of materials, and for an efficient combination / coordination of manned operation and the robot work. Also, configuration of the farm including the arrangement of fields and farm roads, and increase of farm management size to achieve higher cost performance should be considered. These elements of model robot work should reflect the present available state of robot's function and performance, as well as the current state of agricultural sites.