We present a generalised architecture for reactive mobile manipulation while a robot's base is in motion toward the next objective in a high-level task. By performing tasks on-the-move, overall cycle time is reduced compared to methods where the base pauses during manipulation. Reactive control of the manipulator enables grasping objects with unpredictable motion while improving robustness against perception errors, environmental disturbances, and inaccurate robot control compared to open-loop, trajectory-based planning approaches. We present an example implementation of the architecture and investigate the performance on a series of pick and place tasks with both static and dynamic objects and compare the performance to baseline methods. Our method demonstrated a real-world success rate of over 99%, failing in only a single trial from 120 attempts with a physical robot system. The architecture is further demonstrated on other mobile manipulator platforms in simulation. Our approach reduces task time by up to 48%, while also improving reliability, gracefulness, and predictability compared to existing architectures for mobile manipulation. See benburgesslimerick.github.io/ManipulationOnTheMove for supplementary materials.

Grasping is the process of picking up an object by applying forces and torques at a set of contacts. Recent advances in deep learning methods have allowed rapid progress in robotic object grasping. In this systematic review, we surveyed the publications over the last decade, with a particular interest in grasping an object using all six degrees of freedom of the end-effector pose. Our review found four common methodologies for robotic grasping: sampling-based approaches, direct regression, reinforcement learning, and exemplar approaches In addition, we found two “supporting methods” around grasping that use deep learning to support the grasping process, shape approximation, and affordances. We have distilled the publications found in this systematic review (85 papers) into ten key takeaways we consider crucial for future robotic grasping and manipulation research.

This paper introduces DGBench, a fully reproducible open-source testing system to enable benchmarking of dynamic grasping in environments with unpredictable relative motion between robot and object. We use the proposed benchmark to compare several visual perception arrangements. Traditional perception systems developed for static grasping are unable to provide feedback during the final phase of a grasp due to sensor minimum range, occlusion, and a limited field of view. A multi-camera eye-in-hand perception system is presented that has advantages over commonly used camera configurations. We quantitatively evaluate the performance on a real robot with an image-based visual servoing grasp controller and show a significantly improved success rate on a dynamic grasping task.

The DARPA subterranean challenge requires teams of robots to traverse difficult and diverse underground environments. Traversing small gaps is one of the challenging scenarios that robots encounter. Imperfect sensor information makes it difficult for classical navigation methods, where behaviours require significant manual fine tuning. In this paper we present a deep reinforcement learning method for autonomously navigating through small gaps, where contact between the robot and the gap may be required. We first learn a gap behaviour policy to get through small gaps (only centimeters wider than the robot). We then learn a goal-conditioned behaviour selection policy that determines when to activate the gap behaviour policy. We train our policies in simulation and demonstrate their effectiveness with a large tracked robot in simulation and on the real platform. In simulation experiments, our approach achieves 93% success rate when the gap behaviour is activated manually by an operator, and 63% with autonomous activation using the behaviour selection policy. In real robot experiments, our approach achieves a success rate of 73% with manual activation, and 40% with autonomous behaviour selection. While we show the feasibility of our approach in simulation, the difference in performance between simulated and real world scenarios highlight the difficulty of direct sim-to-real transfer for deep reinforcement learning policies. In both the simulated and real world environments alternative methods were unable to traverse the gap.

Legged robots often use separate control policies that are highly engineered for traversing difficult terrain such as stairs, gaps, and steps, where switching between policies is only possible when the robot is in a region that is common to adjacent controllers. Deep Reinforcement Learning (DRL) is a promising alternative to hand-crafted control design, though typically requires the full set of test conditions to be known before training. DRL policies can result in complex (often unrealistic) behaviours that have few or no overlapping regions between adjacent policies, making it difficult to switch behaviours. In this work we develop multiple DRL policies with Curriculum Learning (CL), each that can traverse a single respective terrain condition, while ensuring an overlap between policies. We then train a network for each destination policy that estimates the likelihood of successfully switching from any other policy. We evaluate our switching method on a previously unseen combination of terrain artifacts and show that it performs better than heuristic methods. While our method is trained on individual terrain types, it performs comparably to a Deep Q Network trained on the full set of terrain conditions. This approach allows the development of separate policies in constrained conditions with embedded prior knowledge about each behaviour, that is scalable to any number of behaviours, and prepares DRL methods for applications in the real world.

We quantify the accuracy of various simulators compared to a real world robotic reaching and interaction task. Simulators are used in robotics to design solutions for real world hardware without the need for physical access. The `reality gap' prevents solutions developed or learnt in simulation from performing well, or at all, when transferred to real-world hardware. Making use of a Kinova robotic manipulator and a motion capture system, we record a ground truth enabling comparisons with various simulators, and present quantitative data for various manipulation-oriented robotic tasks. We show the relative strengths and weaknesses of numerous contemporary simulators, highlighting areas of significant discrepancy, and assisting researchers in the field in their selection of appropriate simulators for their use cases.

We present a novel approach of multi-modal deep generative models and apply this to coordinated heterogeneous multi-agent active sensing. A major approach to achieve this objective is to train a multi-modal variational Auto Encoder (M2VAE) that integrates the information of different sensor modalities into a joint latent representation. Furthermore, we derive an objective from the M2VAE that enables the maximization of the evidence lower bound via selection of sensor modalities. Using this approach as a direct reward signal to a multi-modal and multi-agent deep reinforcement learning setup leads intuitively to an epistemic active sensing behavior that coordinately resolves the ambiguity of observations.

Camera viewpoint selection is an important aspect of visual grasp detection, especially in clutter where many occlusions are present. Where other approaches use a static camera position or fixed data collection routines, our Multi-View Picking (MVP) controller uses an active perception approach to choose informative viewpoints based directly on a distribution of grasp pose estimates in real time, reducing uncertainty in the grasp poses caused by clutter and occlusions. In trials of grasping 20 objects from clutter, our MVP controller achieves 80% grasp success, outperforming a single-viewpoint grasp detector by 12%. We also show that our approach is both more accurate and more efficient than approaches which consider multiple fixed viewpoints.

Visual reaching and grasping is a fundamental problem in robotics research. This paper proposes a novel approach based on deep learning a control Lyapunov function and its derivatives by encouraging a differential constraint in addition to vanilla regression that directly regresses independent joint control inputs. A key advantage of the proposed approach is that an estimate of the value of the control Lyapunov function is available in real-time that can be used to monitor the system performance and provide a level of assurance concerning progress towards the goal. The results we obtain demonstrate that the proposed approach is more robust and more reliable than vanilla regression.

We present a deep neural network-based method to perform high-precision, robust and real-time 6 DOF positioning tasks by visual servoing. A convolutional neural network is fine-tuned to estimate the relative pose between the current and desired images and a pose-based visual servoing control law is considered to reach the desired pose. The paper describes how to efficiently and automatically create a dataset used to train the network. We show that this enables the robust handling of various perturbations (occlusions and lighting variations). We then propose the training of a scene-agnostic network by feeding in both the desired and current images into a deep network. The method is validated on a 6 DOF robot.

We present our approach for robotic perception in cluttered scenes that led to winning the recent Amazon Robotics Challenge (ARC) 2017. Next to small objects with shiny and transparent surfaces, the biggest challenge of the 2017 competition was the introduction of unseen categories. In contrast to traditional approaches which require large collections of annotated data and many hours of training, the task here was to obtain a robust perception pipeline with only few minutes of data acquisition and training time. To that end, we present two strategies that we explored. One is a deep metric learning approach that works in three separate steps: semantic-agnostic boundary detection, patch classification and pixel-wise voting. The other is a fully-supervised semantic segmentation approach with efficient dataset collection. We conduct an extensive analysis of the two methods on our ARC 2017 dataset. Interestingly, only few examples of each class are sufficient to fine-tune even very deep convolutional neural networks for this specific task.

The Amazon Robotics Challenge enlisted sixteen teams to each design a pick-and-place robot for autonomous warehousing, addressing development in robotic vision and manipulation. This paper presents the design of our custom-built, cost-effective, Cartesian robot system Cartman, which won first place in the competition finals by stowing 14 (out of 16) and picking all 9 items in 27 minutes, scoring a total of 272 points. We highlight our experience-centred design methodology and key aspects of our system that contributed to our competitiveness. We believe these aspects are crucial to building robust and effective robotic systems.

This paper represents a step towards vision-based manipulation of plastic materials. Manipulating deformable objects is made challenging by: 1) the absence of a model for the object deformation, 2) the inherent difficulty of visual tracking of deformable objects, 3) the difficulty in defining a visual error and 4) the difficulty in generating control inputs to minimise the visual error. We propose a novel representation of the task of manipulating deformable objects. In this preliminary case study, the shaping of kinetic sand, we assume a finite set of actions: pushing, tapping and incising. We consider that these action types affect only a subset of the state, i.e., their effect does not affect the entire state of the system (specialized actions). We report the results of a user study to validate these hypotheses and release the recorded dataset. The actions (pushing, tapping and incising) are clearly adopted during the task, although it is clear that 1) participants use also mixed actions and 2) actions' effects can marginally affect the entire state, requesting a relaxation of our specialized actions hypothesis. Moreover, we compute task errors and corresponding control inputs (in the image space) using image processing. Finally, we show how machine learning can be applied to infer the mapping from error to action on the data extracted from the user study.

The International Joint Conference on Neural Networks (IJCNN) was held in Anchorage (Alaska) in May 2017. This top conference in the field of neural networks included many tracks and special sessions. In particular, a special session on Machine Learning Methods Neural Networks applied to Vision and Robotics (MLMVR) was organized by the authors receiving a large volume of excellent contributions. Only a small set of outstanding papers presented at this special session were invited to submit extended versions of their work. After a rigorous revision process, four of these papers were accepted.

Automated grasping has a long history of research that is increasing due to interest from industry. One grand challenge for robotics is Universal Picking: the ability to robustly grasp a broad variety of objects in diverse environments for applications from warehouses to assembly lines to homes. Although many researchers now openly share code and data, it is challenging to compare and/or reproduce experimental results to identify which aspects of which approaches work best due to variations in assumptions and experimental protocols, e.g., sensors, lighting, robot arms, grippers, and objects.

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. 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.

Robotic vision is limited by line of sight and on-board camera capabilities. Robots can acquire video or images from remote cameras, but processing additional data has a computational burden. This paper applies the Distributed Robotic Vision Service, DRVS, to robot path planning using data outside line-of-sight of the robot. DRVS implements a distributed visual object detection service to distributes the computation to remote camera nodes with processing capabilities. Robots request task-specific object detection from DRVS by specifying a geographic region of interest and object type. The remote camera nodes perform the visual processing and send the high-level object information to the robot. Additionally, DRVS relieves robots of sensor discovery by dynamically distributing object detection requests to remote camera nodes. Tested over two different indoor path planning tasks DRVS showed dramatic reduction in mobile robot compute load and wireless network utilization.

We describe an interactive approach for visual object analysis which exploits the ability of a robot to manipulate its environment. Knowledge of objects' mechanical properties is important in a host of robotics tasks, but their measurement can be impractical due to perceptual or mechanical limitations. By applying a periodic stimulus and matched video filtering and analysis pipeline, we show that even stiff, fragile, or low-texture objects can be distinguished based on their mechanical behaviours. We construct a novel, linear filter exploiting periodicity of the stimulus to reduce noise, enhance contrast, and amplify motion by a selectable gain - the proposed filter is significantly simpler than previous approaches to motion amplification. We further propose a set of statistics based on dense optical flow derived from the filtered video, and demonstrate visual object analysis based on these statistics for objects offering low contrast and limited deflection. Finally, we analyze 7 object types over 59 trials under varying illumination and pose, demonstrating that objects are linearly distinguishable under this approach, and establish the viability of estimating fluid level in a cup from the same statistics.

We propose a system incorporating a tight integration between computer vision and robot control modules on a complex, high-DOF humanoid robot. Its functionality is showcased by having our iCub humanoid robot pick-up objects from a table in front of it. An important feature is that the system can avoid obstacles - other objects detected in the visual stream - while reaching for the intended target object. Our integration also allows for non-static environments, i.e. the reaching is adapted on-the-fly from the visual feedback received, e.g. when an obstacle is moved into the trajectory. Furthermore we show that this system can be used both in autonomous and tele-operation scenarios.

We propose a method for learning specific object representations that can be applied (and reused) in visual detection and identification tasks. A machine learning technique called Cartesian Genetic Programming (CGP) is used to create these models based on a series of images. Our research investigates how manipulation actions might allow for the development of better visual models and therefore better robot vision. This paper describes how visual object representations can be learned and improved by performing object manipulation actions, such as, poke, push and pick-up with a humanoid robot. The improvement can be measured and allows for the robot to select and perform the `right' action, i.e. the action with the best possible improvement of the detector.

Robust object manipulation is still a hard problem in robotics, even more so in high degree-of-freedom (DOF) humanoid robots. To improve performance a closer integration of visual and motor systems is needed. We herein present a novel method for a robot to learn robust detection of its own hands and fingers enabling sensorimotor coordination. It does so solely using its own camera images and does not require any external systems or markers. Our system based on Cartesian Genetic Programming (CGP) allows to evolve programs to perform this image segmentation task in real-time on the real hardware. We show results for a Nao and an iCub humanoid each detecting its own hands and fingers.

In this paper, we present our on-going research to allow humanoid robots to learn spatial perception. We are using artificial neural networks (ANN) to estimate the location of objects in the robot's environment. The method is using only the visual inputs and the joint encoder readings, no camera calibration and information is necessary, nor is a kinematic model. We find that these ANNs can be trained to allow spatial perception in Cartesian (3D) coordinates. These lightweight networks are providing estimates that are comparable to current state of the art approaches and can easily be used together with existing operational space controllers.

To plan complex motions of robots with many degrees of freedom, our novel, very flexible framework builds task-relevant roadmaps (TRMs), using a new sampling-based optimizer called Natural Gradient Inverse Kinematics (NGIK) based on natural evolution strategies (NES). To build TRMs, NGIK iteratively optimizes postures covering task-spaces expressed by arbitrary task-functions, subject to constraints expressed by arbitrary cost-functions, transparently dealing with both hard and soft constraints. TRMs are grown to maximally cover the task-space while minimizing costs. Unlike Jacobian-based methods, our algorithm does not rely on calculation of gradients, making application of the algorithm much simpler. We show how NGIK outperforms recent related sampling algorithms. A video demo (http://youtu.be/N6x2e1Zf_yg) successfully applies TRMs to an iCub humanoid robot with 41 DOF in its upper body, arms, hands, head, and eyes. To our knowledge, no similar methods exhibit such a degree of flexibility in defining movements.

The majority of genetic programming implementations build expressions that only use a single data type. This is in contrast to human engineered programs that typically make use of multiple data types, as this provides the ability to express solutions in a more natural fashion. In this paper, we present a version of Cartesian Genetic Programming that handles multiple data types. We demonstrate that this allows evolution to quickly find competitive, compact, and human readable solutions on multiple classification tasks.

We use a Katana robotic arm to teach an iCub humanoid robot how to perceive the location of the objects it sees. To do this, the Katana positions an object within the shared workspace, and tells the iCub where it has placed it. While the iCub moves it observes the object, and a neural network then learns how to relate its pose and visual inputs to the object location. We show that satisfactory results can be obtained for localisation even in scenarios where the kinematic model is imprecise or not available. Furthermore, we demonstrate that this task can be accomplished safely. For this task we extend our collision avoidance software for the iCub to prevent collisions between multiple, independently controlled, heterogeneous robots in the same workspace.