喜歡就充值下載吧。。。資源目錄里展示的全都有，，下載后全都有,,請放心下載，原稿可自行編輯修改=====================喜歡就充值下載吧。。。資源目錄里展示的全都有，，下載后全都有,,請放心下載，原稿可自行編輯修改=====================喜歡就充值下載吧。。。資源目錄里展示的全都有，，下載后全都有,,請放心下載，原稿可自行編輯修改=====================
本科畢業(yè)論文（設(shè)計）
附錄A外文文獻(xiàn)
Design of an Autonomous Agricultural Robot
Dept. of Industrial Engineering and Management, Ben Gurion University of the Negev
Abstract. This paper presents a state-of-the-art review in the development of autonomous agricultural robots including guidance systems, greenhouse autonomous systems and fruit-harvesting robots. A general concept for a field crops robotic machine to selectively harvest easily bruised fruit and vegetables is designed. Future trends that must be pursued in order to make robots a viable option for agricultural operations are focused upon.
A prototype machine which includes part of this design has been implemented for melon harvesting. The machine consists of a Cartesian manipulator mounted on a mobile chassis pulled by a tractor. Two vision sensors are used to locate the fruit and guide the robotic arm toward it. A gripper grasps the melon and detaches it from the vine. The real-time control hardware architecture consists of a blackboard system, with autonomous modules for sensing, planning and control connected through a PC bus. Approximately 85% of the fruit are successfully located and harvested.
Keywords: Robotics；Autonomous；Agriculture； Intelligent control
1 Introduction
Robots are perceptive machines that can be programmed to perform a variety of agricultural tasks such as cultivating, transplanting, spraying, trimming and selective harvesting. The advent of agricultural robots has the potential of raising the quality of the fresh produce, lowering production costs and reducing the drudgery of manual labor. However, since the agricultural environment is complex and loosely structured fundamental technologies must be developed to solve difficult problems such as: mobile operation in a threedimensional continuously changing track; random location of targets which are difficult to detect and reach (hidden by leaves and positioned among branches); variability in fruit size and shape; delicate products; and hostile environmental conditions like dust, dirt and extreme temperature and humidity. The uncertainties in the fruits locations, size, shape and maturity necessitate a sophisticated sensory system which must identify fruit that are partially occluded in constantly changing illumination conditions (clouds, sun direction) and decide whether a specific fruit is ripe. Uncertainty in location caused by variable vehicle speed and uneven terrain requires velocity and position monitoring. Hence, the overall task requires dynamic, real-time interpretation of the environment and control of various sensing-dependent operations.
The objectives of this paper are to demonstrate the difficulties and complexities involved in the development of autonomous agricultural robots; present the state-of-the-art in development of autonomous agricultural robots; and to layout a general concept and design for an autonomous field crops robotic machine. The initial implementation of this concept for robotic melon harvesting will be presented. The article is summarized with future research and development (R&D) directions which must be pursued before commercial robots can become a viable option for agricultural operations.
2 State-of-the-art Review
2.1 Automatic Guidance Sensors
In contrast to military or public transport robots in which studies of mobile robots are related to constructing a path with wide opportunities of choice according to a complete perception of the environment, automatic guidance of agricultural robots is usually limited to simple problems such as finding the next line of citrus trees to be picked, the next furrow to plough, etc. However, automatic mobile control of agricultural robots is very difficult since the robot operates in a very hostile and unpredictable environment (slopes, hills, mud, rocks). This is further complicated by the fact that the area of movement is relatively large (1 km 2) and although less accurate systems are required (cm vs. mm for industrial operations), the overall system resolution must be quite high to obtain the'necessary accuracies.
A review of automatic guidance sensors for agricultural systems'has been presented by Tillet . This includes mechanical sensing, ultrasonic, radio frequency, gyroscopes, leader cables and optical systems. Several autonomous guidance systems have been developed based on the different sensing approaches: optical techniques for an automated plowing system [4]; photo detectors to detect the furrow ; infrared sensors as indicators for rotary tilling; a vision guided battery powered lawn tractor [12] and positioning sensing systems based on lasers ; passive radar beacons [4] and Geographic Positioning Systems . Microcomputers are used to calculate the tractor position error and different control algorithms e.g., On/Off ; proportional, fuzzy, are used to position and set the correct steering angle. A distributed control system seems to be most suitable for real-time response and was implemented .
2.2 Greenhouse Autonomous Systems
Autonomous operation of vehicles inside the greenhouse is easier since the environment is relatively controllable and more structured than the external agricultural environment. Implementation of AGV's in greenhouses can help reduce hazards of automatic spraying, improve work comfort and labor efficiency and potentially increase accuracy of operations.
Several prototype automatic four-wheel vehicles have been developed for greenhouse transport operations, e.g., infrared guidance; leader cable network routing: In both, the guidance is fully autonomous, however, all tasks are still performed manually. Initial research toward a complete autonomous robot was developed in Japan in which a multipurpose manipulator (tomato picker and selective sprayer) attached to an AGV advances along the rows .
2.3 Summary
The sensing, intelligent control and control elements essential for autonomous control of agricultural robots have been developed. In greenhouses, due to the relatively easier environment, the systems are more advanced and several prototype vehicles have been demonstrated. Prototype fruit harvesters have been developed, however, and the emphasis has been on the critical issue of locating, reaching and picking the fruit and not autonomous guidance. Nevertheless, several algorithms for field guidance and steering control have been developed and implemented. However, only preliminary research has been conducted towards development of a completely autonomous agricultural robot which deals with both automatic vehicle guidance and execution of the agricultural task.
3 Design of an Autonomous Field Crops Robotic Machine
3.1 Sensing
The sensing system must consist of sensors to detect the fruit, determine the fruit ripeness and monitor the robot's location. For robust operation in the unpredictable agricultural environment the robot must be equipped with a redundant multi-sensor system for each of these tasks.
Vehicle location: Continuous update of the vehicle's location is essential for accurate steering of the vehicle along the row and for accurate target interception (since the vehicle advances in an uneven terrain, its velocity is unstable). The true distance must be derived from the path travelled to coordinate between the carrier position at time of target detection and the position at time of intercept.
Vision seems to be the most suitable technique for accurate guidance along the row. Heading error information can be derived by computing the vanishing point of all straight lines generated in the row crop image [6]. A wheeled mobile robot can estimate the current robot point and orientation by accumulating the rotation of the wheels using encoders, gyro compasses, etc. However, due to inherent sensor inaccuracies, the rough surface (which causes slip) and the fact that the estimation errors accumulate as the robot moves an additional position sensor is essential. This sensor can either be a global sensor, e.g., laser, GPS which provides absolute position or a sensor which provides relative location by continuous local sensing of the environment . In both cases, algorithms must be developed to correct the inherent accumulated errors.
Fruit location: Vision sensors tend to be the most suitable technique for dealing with the wide range of sizes, shapes and color of random partially occluded targets. Real-time dedicated imaging hardware is essential for real-time response. Two levels of image sensors are used: one for global path planning (far vision) and a second for local guidance (near vision). To further increase detection reliability and accuracy, different types of sensors must be employed and sensor fusion techniques must be developed to merge information.
Fruit ripeness: must be determined for selective harvesting. Several nondestructive techniques are being explored to evaluate fruit quality (vision, nuclear magnetic resonance, sound, near infra-red) however, most of these systems are not yet available as commercial units. Moreover, once feasible, algorithms must be developed for real-time operation.
3.2 Control
Dedicated and independent controllers must be employed for vehicle guidance, manipulator motion control, gripper and sensor activities.
Vehicle guidance: This control system must actuate the vehicle steering and control the overall vehicle motion: stop-start; forward speed, etc.
Manipulator control: is responsible for exact position control of each actuator with closed-loop integrated feedback.
Gripper control: the gripper controller modifies the wrist and grasp motions.
Sensors control: since global detection devices are not sufficiently accurate for complete control of the tasks, actuators are often necessary to complete the sensing operation. Therefore, active sensing must be employed e.g., directing a blower towards a suspected fruit location, snapping additional images at a different angle.
3.3 Intelligent Control System and Algorithms
Sensor-based control applications require a flexible, real-time multitasking and parallel programming environment which integrate perception, planning and control. Therefore, the intelligent control system, was designed as distributed with independent sensing, planning and control modules [10]. The blackboard approach [11] was applied since it is useful for complex and ill structured problems and uses opportunistic reasoning. This is important for robust operation in the dynamic agricultural environment where for example, changes in light conditions because of clouds or shades require different image processing routines; and fruit distribution variations imply activating different motion control algorithms. For each part of the problem and at each stage of the solution, the best knowledge representation and solution strategy can be selected. Qualitative and quantitive methods can be combined.
The agricultural robot blackboard system (Figure 1) included as the knowledge sources data from two vision sensors (Far and Near), distance encoder's data, vehicle's location and information about the current robot/gripper state provided by the robot sensors. The control modules are represented by several separate processes running in parallel and controlling different hardware devices such as the robotic motors, vehicle steering and control. To reduce complexity, the intelligent controller is divided into two hierarchies: a high-level planning blackboard and a low-level control blackboard.
The vision system must be a stand alone independent module due to its complexity and since it provides the most important source of information about the robot's environment. Dedicated real-time image processing hardware and software must be used. A hierarchy of sensors is introduced for the planning (far vision) and control (near vision) levels. Additional sensing modules must be incorporated for fruit detection, fruit ripeness determination and for vehicle guidance.
A dedicated motion control board must handle all the encoder input decoding, digital filtering of the control signals and generation of analog motor commands relate to the manipulator control. In addition, the motion controller must provide for limit and emergency stop switches. Additional dedicated motor control units must be employed for steering the vehicle and issuing movement and speed commands.
Fig. 1. Schematic architecture of the blackboard robot controller.
3.4 Implementation of a Robotic Melon Harvester
System Design
The robotic melon harvester consists of a robot arm mounted on a mobile platform which is drawn by a tractor (Figure 2). The platform is a rectangular steel
frame with two wheels attached to it at the rear end. The front end is attached to the drawbar of the tractor. The platform is divided into two areas. The front area contains one of the visual sensors (far camera) and provides a clear field of view of the whole bed for detection of oncoming melons. Blowers mounted on the platform clear the foliage from the melons to expose the melons which are hidden by the canopy.
A Cartesian manipulator is located in the second area, right behind the first one. The manipulator's work space is 1.5 by 1.5 m. This enables access to any location of the melon across the bed and to the special conveyers positioned along the platform on both sides. The vertical clearance of the gripper arm is 0.9 m which is enough to pick a fruit and load it into the conveyers located along both sides of the platform. A pneumatic gripper is attached to the robot arm by a flexible joint to absorb side loads caused by horizontal motions during fruit picking. The manipulator is powered by electrical DC servo motors with encoder feedback. A second (near) visual sensor is mounted in the picking area on the gripper itself at a constant height. It can be considered as an-eye-on-the-hand type of architecture that provides final information to direct the picking mechanism. The near camera is located at a smaller height and has a narrow field of view.
Fig. 2. Schematic design of the prototype robotic melon harvester,
a. Side view; b. Front View; c. Top view; d. Photograph.
Motor Control: The motors are connected to signal amplifiers which execute servo-motor control according to software. For interfacing with the amplifiers a PC based motion control interface card is used (Galil DMC-600). The position of the manipulator arm along its three axes is controlled by a computer program, so tt can be sent to any desired position within the track boundaries. To prevent over running in case of program failure, proximity sensors were installed at both ends of all three axes to limit travel. An emergency stop signal is automatically generated and all motors are switched off if the arm reaches one of these sensors. To avoid collision into the ground, an additional sensor was connected to the Z axis. If the ground (or any other obstacle) is reached during a vertical downward movement, the emergency stop signal is also generated. In addition, the DMC card has one 8-bit I/O port which is used for tractor motion control (stop-start) and the gripper/cutter control.
Vehicle motion: An incremental optical encoder (Renco, R-250) was used to determine the path traveled by the carrier. The encoder produces 256 pulses per revolution. These pulses are the input for a special hardware decoding card. The output of the decoding scheme is a byte which has the value equal to the counted pulses since last decoder reset was done. The reset is used for setting the counter's value to zero. The accuracy of the decoder is 0.92 pulses/cm. The decoder output byte was transmitted to the PC via the standard parallel printer port.
The tractor, attached to the robot platform is moved by a special hydraulic motor which provides a constant speed. The motor was connected so as to provide capability to move the robot both forward and backward. The movement direction is set up by a special electrical switch which has a control input line connected to the output port of the DMC controller. This enables control of the hydraulic motor by the control software. The program can issue three motion commands Forward, Backward and Stop to activate or stop the robot movement. Steering correction was provided by a human operator.
4 Conclusions and Future Trends
A robotic field crops robotic machine has been designed to selectively harvest easily bruised fruit and vegetables. The overall design could be easily adapted to facilitate additional tasks such as harvesting similar crops, selective spraying, and transplanting. A prototype machine to harvest melons has been constructed and field tested.
Most sensory systems and algorithms for fruit detection are capable of detecting between 80-85%. Multiple sensors must be used to increase detection reliability and accuracy. By employing sensor fusion techniques that merge a multitude of information, more complete knowledge of the scene can be obtained. Strategies to determine when and which sensor should be employed and how to combine the multitude of information must be formulated and implemented [2]. Active sensing could further improve performance.
Two technologies are lacking for the complete development of a mobile selective harvester: ripeness determination and 3-D dynamic robotic guidance. The ripeness sensor is fruit dependent and therefore will probably be the limiting factor since individual R&D must be directed for each specific application.
Another issue that must be investigated is the control of the robot arm to approach and grasp the fruit while the chassis is in motion. Since the chassis is moving in a constantly changing terrain, algorithms must be developed for object tracking under uncertainties. Preliminary algorithms have been developed and implemented in real-time based on the superposition method, a mathematical model of strategies that humans use when modifying ongoing motions .
The successful development of these issues will lead to the development of a mobile, autonomous field robotic machine. This will undoubtedly be a breakthrough towards mechanization of highly perishable agricultural products and improve agricultural operations. Economic feasibility of such a system will be enhanced by the possibility of multiple uses of the robot such as application for harvesting other crops and perform additional tasks such as transplanting. Addition of post harvest tasks like sizing and sorting will also increase this economic potential.
There is no consensus on the viability of agricultural robots as an alternative method for manual operations : while all will agree that no cost-effective product is yet available on the market for fully independent operation, some will argue that with the rapid developments in computers and sensors, it is only a matter of time and money; while others still maintain that agricultural robots will never be economically practical. However, with the increasing costs of labour, demand for high quality fresh produce, and decreasing costs of computers on one-hand with increasing power on the other hand, the break even point might be closer than expected. Nevertheless, all agree that R&D in this area is exciting and challenging.
外文翻譯
自主農(nóng)業(yè)機器人設(shè)計
工業(yè)工程與管理部門，內(nèi)蓋夫本古里安大學(xué)
摘要；本文對自主農(nóng)業(yè)機器人的發(fā)展現(xiàn)狀進(jìn)行了綜述，包括制導(dǎo)系統(tǒng)、溫室自治系統(tǒng)。設(shè)計了一種可以選擇性地收獲容易被碰傷的水果和蔬菜的野外作物機器人機器的一種概念。為了使機器人成為農(nóng)業(yè)生產(chǎn)的可行選擇，未來對機器人趨勢必須加以關(guān)注。
一個原型機的部分設(shè)計已經(jīng)實現(xiàn)了甜瓜收獲。這臺機器由一輛由拖拉機牽引的移動底盤上的笛卡爾機械手組成。兩個視覺傳感器被用來定位水果和引導(dǎo)機器人手臂朝向它。一個夾子抓住甜瓜，把它從藤上分離出來。實時控制硬件體系結(jié)構(gòu)由黑板系統(tǒng)組成，通過PC總線進(jìn)行傳感、規(guī)劃和控制的自主模塊。大約85%的果實被成功地定位和收獲。
關(guān)鍵詞：機器人；自主；農(nóng)業(yè)；智能控制。
1簡介
機器人是一種感知機，可以編程完成各種農(nóng)業(yè)任務(wù)，如培養(yǎng)、移植、噴灑、修剪和選擇性收割。農(nóng)業(yè)機器人的出現(xiàn)有可能提高新鮮農(nóng)產(chǎn)品的質(zhì)量，降低生產(chǎn)成本，減少手工勞動的苦工。然而，由于農(nóng)業(yè)環(huán)境是復(fù)雜的、結(jié)構(gòu)松散的基礎(chǔ)技術(shù)，必須發(fā)展起來，以解決諸如:在三維不斷變化的軌道上移動操作的難題;目標(biāo)的隨機位置，很難探測和到達(dá)(隱藏在樹葉和樹枝之間);水果大小和形狀的變化;奇特的產(chǎn)品;以及惡劣的環(huán)境條件，如灰塵，污垢和極端的溫度和濕度。
水果的位置、大小、形狀和成熟度的不確定性，需要一個復(fù)雜的感官系統(tǒng)，它必須識別部分遮擋在不斷變化的光照條件(云層，太陽方向)的水果，并決