Search

Many hierarchical reinforcement learning algorithms utilise a series of independent skills as a basis to solve tasks at a higher level of reasoning. These algorithms don't consider the value of using skills that are cooperative instead of independent. This paper proposes the Cooperative Consecutive Policies (CCP) method of enabling consecutive agents to cooperatively solve long time horizon multi-stage tasks. This method is achieved by modifying the policy of each agent to maximise both the current and next agent's critic. Cooperatively maximising critics allows each agent to take actions that are beneficial for its task as well as subsequent tasks. Using this method in a multi-room maze domain and a peg in hole manipulation domain, the cooperative policies were able to outperform a set of naive policies, a single agent trained across the entire domain, as well as another sequential HRL algorithm.

The rise of herbicide resistant weed species has reinvigorated research in nonchemical methods for weed management. Robots, such as AgBot II, that can detect and classify weeds as they traverse a field are a key enabling factor for individualised treatment of weed species. Integral to the invidualized treatment of weed species are the the nonherbicide methods through which the weeds are managed. This letter explores mechanical methods as an alternative to weed management. Three implements are considered: below-surface tilling (arrow hoe), above-surface tilling (tines), and a cutting mechanism. These mechanisms were evaluated in a controlled field with varying rates of application to herbicide-resistant weeds of interest for Queensland, Australia. Statistical analysis demonstrated the efficacy of these implements and highlighted the importance of early intervention. It was found that a tine, deployed automatically on AgBot II, was effective for all of the weeds considered in this study, leading to an overall survival probability of 0.28 ± 0.15. Further analysis demonstrated the significance of treatment time with late intervention commencing at week 6 resulting in a survival probability of 0.54 ± 0.08 vs 0.24 ± 0.18 for earlier intervention at week 4.

Robotic challenges like the Amazon Picking Challenge (APC) or the DARPA Challenges are an established and important way to drive scientific progress. They make research comparable on a well-defined benchmark with equal test conditions for all participants. However, such challenge events occur only occasionally, are limited to a small number of contestants, and the test conditions are very difficult to replicate after the main event. We present a new physical benchmark challenge for robotic picking: the ACRV Picking Benchmark (APB). Designed to be reproducible, it consists of a set of 42 common objects, a widely available shelf, and exact guidelines for object arrangement using stencils. A well-defined evaluation protocol enables the comparison of complete robotic systems -- including perception and manipulation -- instead of sub-systems only. Our paper also describes and reports results achieved by an open baseline system based on a Baxter robot.

In this letter, we present a new robotic harvester (Harvey) that can autonomously harvest sweet pepper in protected cropping environments. Our approach combines effective vision algorithms with a novel end-effector design to enable successful harvesting of sweet peppers. Initial field trials in protected cropping environments, with two cultivar, demonstrate the efficacy of this approach achieving a 46% success rate for unmodified crop, and 58% for modified crop. Furthermore, for the more favourable cultivar we were also able to detach 90% of sweet peppers, indicating that improvements in the grasping success rate would result in greatly improved harvesting performance.

The rapid evolution of herbicide-resistant weed species has revitalized research in nonchemical methods for weed destruction. Robots with vision-based capabilities for online weed detection and classification are a key enabling factor for the specialized treatment of individual weed species. This paper describes the design, development, and testing of a modular robotic platform with a heterogeneous weeding array for agriculture. Starting from requirements derived from farmer insights, technical specifications are put forward. A design of a robotic platform is conducted based on the required technical specifications, and a prototype is manufactured and tested. The second part of the paper focuses on the weeding mechanism attached to the robotic platform. This includes aspects of vision for weed detection and classification, as well as the design of a weeding array that combines chemical and mechanical methods for weed destruction. Field trials of the weed detection and classification system show an accuracy of 92.3% across a range of weed species, while the heterogeneous weed management system is able to selectively apply a mechanical or chemical control method based on the species of weed. Together, the robotic platform and weeding array demonstrate the potential for robotic plant-species–specific weed management enabled by the vision-based online detection and classification algorithms.

This paper presents a 3D visual detection method for the challenging task of detecting peduncles of sweet peppers (Capsicum annuum) in the field. Cutting the peduncle cleanly is one of the most difficult stages of the harvesting process, where the peduncle is the part of the crop that attaches it to the main stem of the plant. Accurate peduncle detection in 3D space is therefore a vital step in reliable autonomous harvesting of sweet peppers, as this can lead to precise cutting while avoiding damage to the surrounding plant. This paper makes use of both colour and geometry information acquired from an RGB-D sensor and utilises a supervised-learning approach for the peduncle detection task. The performance of the proposed method is demonstrated and evaluated using qualitative and quantitative results (the Area-Under-the-Curve (AUC) of the detection precision-recall curve). We are able to achieve an AUC of 0.71 for peduncle detection on field-grown sweet peppers. We release a set of manually annotated 3D sweet pepper and peduncle images to assist the research community in performing further research on this topic.

Autonomously positioning a robot arm toolpoint, mobile base or UAV relative to a surface of interest is a fundamental capability in many applications such as infrastructure monitoring and planetary rock inspection. Robots performing this task in uncontrolled real world environments face many challenges such as adverse weather conditions, changing illumination or even the need to perform positioning using multiple sensing modalities. In this paper, we address this key robotics challenge in the context of the Mars2020 rover mission. Scientists require the ability to manually identify a target on a rocky outcrop using a centrally mounted RGB sensor and command the rover to autonomously position a robot arm toolpoint at that target on the following day, relying only on the x-ray sensor mounted on the arm toolpoint itself. We use an adapted sequence-based technique for multi-modal image registration that builds on recent appearance-invariant robotic place recognition algorithms. We introduce a new large, custom dataset of rock samples with multimodal sensing observations, and compare the performance of the proposed technique to a convolutional neural network (CNN)-based approach as well as a traditional Speeded-Up Robust Features (SURF)-based approach. Finally, we demonstrate the entire system performing vision-based positioning of a robot arm tool point on actual rock samples with significant illumination change.

This paper presents a method for estimating the 6DOF pose of sweet-pepper (capsicum) crops for autonomous harvesting via a robotic manipulator. The method uses the Kinect Fusion algorithm to robustly fuse RGB-D data from an eye-in-hand camera combined with a colour segmentation and clustering step to extract an accurate representation of the crop. The 6DOF pose of the sweet peppers is then estimated via a nonlinear least squares optimisation by fitting a superellipsoid to the segmented sweet pepper. The performance of the method is demonstrated on a real 6DOF manipulator with a custom gripper. The method is shown to estimate the 6DOF pose successfully enabling the manipulator to grasp sweet peppers for a range of different orientations. The results obtained improve largely on the performance of grasping when compared to a naive approach, which does not estimate the orientation of the crop.

This paper presents a novel crop detection system applied to the challenging task of field sweet pepper (capsicum) detection. The field-grown sweet pepper crop presents several challenges for robotic systems such as the high degree of occlusion and the fact that the crop can have a similar colour to the background (green on green). To overcome these issues, we propose a two-stage system that performs per-pixel segmentation followed by region detection. The output of the segmentation is used to search for highly probable regions and declares these to be sweet pepper. We propose the novel use of the local binary pattern (LBP) to perform crop segmentation. This feature improves the accuracy of crop segmentation from an AUC of 0.10, for previously proposed features, to 0.56. Using the LBP feature as the basis for our two-stage algorithm, we are able to detect 69.2% of field grown sweet peppers in three sites. This is an impressive result given that the average detection accuracy of people viewing the same colour imagery is 66.8%.

Natural language provides a convenient means of communicating information, and as such, is an ideal medium for enabling non-expert users to teach robots novel tasks. However in order to take advantage of natural language, a series of challenges must first be overcome. These challenges include the need to - a) generalise learnt tasks to novel scenarios without retraining, - b) resolve problems encountered during task execution, and - c) derive implicit information from knowledge about the domain. To solve these challenges, this paper presents a novel approach to learning complex hierarchical tasks through natural language instruction, which not only allows learnt tasks to be generalised to novel situations without the need for retraining, but also enables an agent to derive implicit information from domain knowledge. Additionally, the approach presented in this paper enables the agent to infer task properties, such as preconditions and effects, directly from the explanation of the task flow. We validate our approach by demonstrating an implementation of our algorithms both on a simulated agent, as well as a Baxter robot. In each case, the agent is provided with a small set of primitive tasks for manipulating its workspace. From these primitives, we demonstrate the ability to teach the agent increasing complex tasks, including tasks of table cleaning, solely through natural language instructions.

This paper proposes new metrics and a performance-assessment framework for vision-based weed and fruit detection and classification algorithms. In order to compare algorithms, and make a decision on which one to use fora particular application, it is necessary to take into account that the performance obtained in a series of tests is subject to uncertainty. Such characterisation of uncertainty seems not to be captured by the performance metrics currently reported in the literature. Therefore, we pose the problem as a general problem of scientific inference, which arises out of incomplete information, and propose as a metric of performance the(posterior) predictive probabilities that the algorithms will provide a correct outcome for target and background detection. We detail the framework through which these predicted probabilities can be obtained, which is Bayesian in nature. As an illustration example, we apply the framework to the assessment of performance of four algorithms that could potentially be used in the detection of capsicums (peppers).

This paper presents a global-optimisation frame-work for the design of a manipulator for harvesting capsicum(peppers) in the field. The framework uses a simulated capsicum scenario with automatically generated robot models based on DH parameters. Each automatically generated robot model is then placed in the simulated capsicum scenario and the ability of the robot model to get to several goals (capsicum with varying orientations and positions) is rated using two criteria:the length of a collision-free path and the dexterity of the end-effector. These criteria form the basis of the objective function used to perform a global optimisation. The paper shows a preliminary analysis and results that demonstrate the potential of this method to choose suitable robot models with varying degrees of freedom.

This thesis develops a novel approach to robot control that learns to account for a robot's dynamic complexities while executing various control tasks using inspiration from biological sensorimotor control and machine learning. A robot that can learn its own control system can account for complex situations and adapt to changes in control conditions to maximise its performance and reliability in the real world. This research has developed two novel learning methods, with the aim of solving issues with learning control of non-rigid robots that incorporate additional dynamic complexities. The new learning control system was evaluated on a real three degree-of-freedom elastic joint robot arm with a number of experiments: initially validating the learning method and testing its ability to generalise to new tasks, then evaluating the system during a learning control task requiring continuous online model adaptation.

Developing accurate and reliable crop detection algorithms is an important step for harvesting automation in horticulture. This paper presents a novel approach to visual detection of highly-occluded fruits. We use a conditional random field (CRF) on multi-spectral image data (colour and Near-Infrared Reflectance, NIR) to model two classes: crop and background. To describe these two classes, we explore a range of visual-texture features including local binary pattern, histogram of oriented gradients, and learn auto-encoder features. The pro-posed methods are evaluated using hand-labelled images from a dataset captured on a commercial capsicum farm. Experimental results are presented, and performance is evaluated in terms of the Area Under the Curve (AUC) of the precision-recall curves.Our current results achieve a maximum performance of 0.81AUC when combining all of the texture features in conjunction with colour information.