“Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks”の学習報告

This manuscript describes a unique class of locomotive robot: A soft robot, composed exclusively of soft materials (elastomeric polymers), which is inspired by animals (e.g., squid, starfish, worms) that do not have hard internal skeletons. Soft lithography was used to fabricate a pneumatically actuated robot capable of sophisticated locomotion (e.g., fluid movement of limbs and multiple gaits). This robot is quadrupedal; it uses no sensors, only five actuators, and a simple pneumatic valving system that operates at low pressures (< 10 psi). A combination of crawling and undulation gaits allowed this robot to navigate a difficult obstacle. This demonstration illustrates an advantage of soft robotics: They are systems in which simple types of actuation produce complex motion.

Multigait soft robot

http://bme2.aut.ac.ir/~towhidkhah/MPC/seminars-ppt/seminar%20MPC/Fazane%20karami/Reference/Architectures%20for%20distributed%20and%20hierarchical%20Model%20Predictive%20Control%20%e2%80%93%20A%20review.pdf

Architectures for distributed and hierarchical Model Predictive Control - A review

http://pdclab.seas.ucla.edu/Publications/PDChristofides/PDChristofides_RScattolini_DMunozdelaPena_JLiu_CCE_2013_21_DMPC_A_Tutorial_Review_and_Future_Research_Directions.pdf

Distributed model predictive control: A tutorial review and future research directions

This paper presents a novel approach to fruit detection using deep convolutional neural networks. The aim is to build an accurate, fast and reliable fruit detection system, which is a vital element of an autonomous agricultural robotic platform; it is a key element for fruit yield estimation and automated harvesting. Recent work in deep neural networks has led to the development of a state-of-the-art object detector termed Faster Region-based CNN (Faster R-CNN). We adapt this model, through transfer learning, for the task of fruit detection using imagery obtained from two modalities: colour (RGB) and Near-Infrared (NIR). Early and late fusion methods are explored for combining the multi-modal (RGB and NIR) information. This leads to a novel multi-modal Faster R-CNN model, which achieves state-of-the-art results compared to prior work with the F1 score, which takes into account both precision and recall performances improving from 0 . 807 to 0 . 838 for the detection of sweet pepper. In addition to improved accuracy, this approach is also much quicker to deploy for new fruits, as it requires bounding box annotation rather than pixel-level annotation (annotating bounding boxes is approximately an order of magnitude quicker to perform). The model is retrained to perform the detection of seven fruits, with the entire process taking four hours to annotate and train the new model per fruit.

DeepFruits: A Fruit Detection System Using Deep Neural Networks.

This paper describes the concept of an autonomous robot for harvesting cucumbers in greenhouses. A description is given of the working environment of the robot and the logistics of harvesting. It is stated that for a 2 ha Dutch nursery, 4 harvesting robots and one docking station are needed during the peak season. Based on these preliminaries, the design specifications of the harvest robot are defined. The main requirement is that a single harvest operation may take at most 10 s. Then, the paper focuses on the individual hardware and software components of the robot. These include, the autonomous vehicle, the manipulator, the end-effector, the two computer vision systems for detection and 3D imaging of the fruit and the environment and, finally, a control scheme that generates collision-free motions for the manipulator during harvesting. The manipulator has seven degrees-of-freedom (DOF). This is sufficient for the harvesting task. The end-effector is designed such that it handles the soft fruit without loss of quality. The thermal cutting device included in the end-effector prevents the spreading of viruses through the greenhouse. The computer vision system is able to detect more than 95% of the cucumbers in a greenhouse. Using geometric models the ripeness of the cucumbers is determined. A motion planner based on the Aa-search algorithm assures collision-free eye-hand co-ordination. In autumn 2001 system integration took place and the harvesting robot was tested in a greenhouse. With a success rate of 80%, field tests confirmed the ability of the robot to pick cucumbers without human interference. On average the robot needed 45 s to pick one cucumber. Future research focuses on hardware and software solutions to improve the picking speed and accuracy of the eye-hand co-ordination of the robot.

An Autonomous Robot for Harvesting Cucumbers in Greenhouses

Since both the spatial and vertical heterogeneities in soil properties have an impact on crop growth and yield, accurate characterization of soil properties at high sampling resolution is a preliminary step in successful management of soil-water-plant system. Conventional soil sampling and analyses have shown mixed economical returns due to the high costs associated with labor-intensive sampling and analysis procedures, which might be accompanied with map uncertainties. Therefore, the conventional laboratory methods are being replaced or complemented with the analytical soil sensing techniques. The objective of this chapter is to review different soil sensing methods used to characterize key soil properties for management of soil-water-plant system. This will cover laboratory, in situ in the field, and on-line measurement methods. This review chapter is furnished with an overview of background information about a sensing concept, basic principle and brief theory, various factors affecting the output of the sensor, and justification of why specific soil properties can be related with its output. The literature review is succeeded with an integration and analysis of findings in view of application in the precision agriculture domain. Potentials and limitations of current sensor technologies are discussed and compared with commonly used state-of-the-art laboratory techniques. As sensing is commonly addressed as a very technical discipline, the match between the information currently collected with sensors and those required for site-specific application of different inputs, and crop growth and development is discussed, highlighting the most accurate method to measure a soil property for a given application.

Sensing soil properties in the laboratory, in situ, and on-Line: A review

Field Test of an Autonomous Cucumber Picking Robot

In this thesis a methodology is developed for the construction and analysis of an optimal greenhouse climate control system. In chapter 1, the results of a literature survey are presented and the research objectives are defined. In the literature, optimal greenhouse climate management systems have been commonly presented and analysed as hierarchical systems. The main reasons were the inherent complexity of the system and existing differences in dynamic response times of the process variables involved. In general terms, the hierarchical control schemes contained two layers. An upper layer emphasizes control of the crop growth dynamics and a lower layer is concerned with control of the greenhouse climate. With respect to optimal greenhouse climate management it has not become completely clear from the literature how the hierarchical structure should be derived, how the relations between the different layers should be defined nor what kind of performance criterion should be used at each level. It is one of the objectives of this thesis to investigate the hierarchical decomposition of greenhouse climate management. Until now, in research on optimal greenhouse climate management, the main emphasis was on control of the carbon dioxide concentration and temperature in the greenhouse. Since in horticultural practice as well as in research, the humidity level in the greenhouse is considered to be an important climate variable determining the quality and quantity of crop production, a second objective of this research is to explicitly include humidity control in the control system design. For application of optimal control in horticultural practice, it is necessary to have a quantative model which describes the dynamic response of the greenhouse crop production process to the control and exogenous inputs with some accuracy. A third objective of this thesis is to analyse and validate dynamic models of the greenhouse crop production process. In optimal greenhouse climate control, predictions of the auction price and the weather are important. A final objective of this thesis is to investigate briefly the predictability of auction prices and to develop a methodology for greenhouse climate control which deals with errors in the model and the predictions of both auction price and weather, while maintaining near optimal performance. In this thesis the production of a lettuce crop is used as a vehicle for the illustration of the methodology developed. In chapter 2 the optimal greenhouse climate control problem is defined. The objective of optimal greenhouse climate control during the production of a lettuce crop is defined as to control the greenhouse climate such that the net return is maximized. The net return is defined as the difference between the revenues of the harvested lettuce crop and the climate conditioning costs. In the control problem physical limitations of the control inputs are included. Also simple bound constraints are imposed on the state of the greenhouse climate. These constraints represent unmodelled effects of the greenhouse climate on the quality and quantity of crop production. In chapter 3, dynamic models of crop growth and greenhouse climate are presented and analysed. With data collected in two lettuce growth experiments in a greenhouse, the models used in this research are calibrated and validated. It is shown that the models quite accurately describe the dynamics of the process variables considered in this research. Also it is illustrated that in the crop production process differences in response times do exist. The greenhouse climate responds rapidly to changes in control and external inputs, whereas crop growth responds comparatively slow to changes in the environmental conditions. A methodology for a first-order sensitivity analysis of dynamic models is presented and one of the lettuce growth models is analysed. It is shown that in this model of lettuce growth, only a few parameters mainly determine crop growth. The solar radiation level and carbon dioxide concentration are important environmental conditions. The temperature is relatively less important in this model. At the end of the chapter the models of lettuce growth and greenhouse climate are integrated to yield a model of the crop production process. This model is also validated. In chapter 4, the performance criterion used in optimal greenhouse climate control is defined in more detail. It is shown that a linear relation between the harvest fresh weight of the crop and the auction price of lettuce exists. The time-varying parameters of this linear relation show distinct diurnal trends which offer an opportunity for predicting the auction price. In chapter 5, the methodology for the solution and analysis of optimal control problems is presented. Necessary conditions for the existence of an optimal control trajectory are derived for a control problem with state and control constraints. It is shown that, in optimal control, the necessary conditions for optimality have a meaningful economic interpretation which can be used in the analysis of optimal greenhouse climate management. Considerable attention is paid to the solution of control problems in which both slow and fast dynamics interact. Using the framework of singular perturbed systems, the control problem in which both slow and fast dynamics are included, is decomposed into two subproblems. One sub-problem emphasizes efficient control of the slow dynamics, the other sub-problem emphasizes efficient control of the fast dynamics. It is shown that the performance criteria used in both sub-problems have a clear relationship with the main objective of the original control problem. The methodology of a first-order sensitivity analysis of openloop optimal control problems is derived. An algorithm is presented which is used to solve the optimal control problem iteratively on a digital computer is described. Finally, a feedback, feedforward control scheme based on the framework of optimal control is derived which is expected to yield near optimal performance in the presence of perturbations in the model and in the external inputs. In chapter 6 the methodology derived in chapter 5 is applied to the optimal greenhouse climate control problem using the model defined in chapter 3 and the performance criterion defined in chapter 4. For the full control problem, in which both greenhouse climate dynamics and crop growth dynamics are included, the necessary conditions are derived. Also the equations of the slow sub-problem, aiming at economic optimal control of the crop growth dynamics, and of the fast sub-problem, emphasizing efficient control of the greenhouse climate dynamics, are explicitly stated. The equations are analysed to gain insight into the operation of optimal greenhouse climate control. In chapter 7, optimal greenhouse climate control is investigated with simulations. Using the measured data obtained during one of the lettuce growth experiments in the greenhouse, optimal greenhouse climate control is compared with climate control supervised by the grower. The simulations show that using optimal control, greenhouse heating, carbon dioxide supply and ventilation are more efficiently used. Also, it is shown that the humidity level in the greenhouse largely determines the ventilation rate and strongly affects the performance of optimal greenhouse climate control. The feedback, feedforward control scheme is evaluated in simulations. It is shown that using a short-term weather prediction, the control scheme is able to achieve near optimal performance in the presence of large perturbations of the state and the weather from precalculated nominal trajectories. These results suggest that accurate long term weather predictions no longer limit the application of optimal greenhouse climate control in practice. In simulations, the decomposition of the greenhouse climate control problem is successfully evaluated. The performance of both sub-problems closely approximates the performance of the full control problem in which both slow and fast dynamics are included. Also, the control trajectories resulting from the decomposition closely approximate the control trajectories calculated for the full problem. Another simulation shows the importance of using an explicit economic criterion for control of the greenhouse climate dynamics. The decomposition supplies valuable information on the form of the objective function appropriate for use in control of the greenhouse climate dynamics. Finally, a sensitivity analysis supplies insight into the performance sensitivity of open-loop optimal greenhouse climate for perturbations in the model parameters and initial conditions. Chapter 8 contains a synthesis of the results obtained in the previous chapters. Two concepts for the implementation of optimal greenhouse climate management in horticultural practice are presented. The concepts are compared and their operation is discussed. It is indicated that both control systems are able to deal with uncertainty in the weather and the auction prices. Also, practical aspects like the amount of CPU- time needed, the flexibility of the hierarchical structure, the role of the grower in greenhouse climate management as well as the contribution of optimal control to the improvement of the efficiency of greenhouse crop production are discussed. Finally, it is indicated how the methodology developed in this thesis can be applied to multiple harvest crops like tomatoes and cucumbers. Chapter 9 contains concluding remarks as well as suggestions for further research.

/pdf/greenhouse-climate-management-an-optimal-control-approach-1vlmgknssx.pdf

Greenhouse climate management : an optimal control approach

Features Discusses economic optimization of greenhouse control through mathematical modeling Examines 30 years of scientific research to present a unified framework for efficient decision-making Presents modern methods of control and optimization including classical rule-based and multivariable feedback controllers Utilizes real and experimental examples and novel case discussions such as solar greenhouses Concludes with a discussion of open issues to stimulate new areas of research and development Summary Greenhouse control system manufacturers produce equipment and software with hundreds of settings and, while they hold training courses on how to adjust these settings, there is as yet no integrated instruction on when or why. Despite rapid growth in the greenhouse industry, growers are still faced with a multitude of variables and no unifying framework from which to choose the best option. Consolidating 30 years of research in greenhouse climate control, Optimal Control of Greenhouse Cultivation utilizes mathmatical models to incorporate the wealth of scientific knowledge into a feasible optimal control methodology for greenhouse crop cultivation. Discussing several different paradigms on greenhouse climate control, it integrates the current research into physical modeling of the greenhouse climate in response to heating, ventilation, and other control variables with the biological modeling of variables such as plant evapo-transpiration and growth. Key topics include state-space greenhouse and crop modeling needed for the design of integrated optimal controllers that exploit rather than mitigate outside weather conditions, especially sunlight, given widely different time scales. The book reviews classical rule-based and multivariable feedback controllers in comparison with the optimal hierarchical control paradigm. It considers real and hypothetical examples including lettuce, tomato, and solar greenhouses and examines experimental results of greenhouse climate control using optimal control software. The book concludes with a discussion of open issues as well as future perspectives and challenges. Providing a tool to automatically determine the most economical controls and settings for their operation, this much-needed book relieves growers of unnecessary control tasks, and allows them to achieve the best possible trade-off between short term savings and optimal harvest yield.

E.J. van Henten

Papers

An Autonomous Robot for Harvesting Cucumbers in Greenhouses

Sensing soil properties in the laboratory, in situ, and on-Line: A review

Field Test of an Autonomous Cucumber Picking Robot

Greenhouse climate management : an optimal control approach

Optimal control of greenhouse cultivation