ICRA 2018 Papers

Driven by the promise of leveraging the benefits of collaborative robot operation, this paper presents an approach to estimate the relative transformation between two small Unmanned Aerial Vehicles (UAVs), each equipped with a single camera and an inertial sensor, comprising the first step of any meaningful collaboration. Formation flying and collaborative object manipulation are some of the few tasks that the proposed work has direct applications on, while forming a variable-baseline stereo rig using two UAVs carrying a monocular camera each promises unprecedented effectiveness in collaborative scene estimation. Assuming an overlap in the UAVs' fields of view, in the proposed framework, each UAV runs monocular-inertial odometry onboard, while an Extended Kalman Filter fuses the UAVs' estimates and common image measurements to estimate the metrically scaled relative transformation between them, in realtime. Decoupling the direction of the baseline between the cameras of the two UAVs from its magnitude, this work enables consistent and robust estimation of the uncertainty of the relative pose estimation. Our evaluation on both on simulated data and benchmarking datasets consisting of real aerial data, reveals the power of the proposed methodology in a variety of scenarios. Video - https://youtu.be/AmkkaXa2601.

Intelligent vehicles require an efficient way to compute a feasible path that connects its current localization to a destination point. To achieve that, the knowledge about the road network, including its geometry, is important since connections between roads can be used in the path planning. This work consists on computing a feasible global path for autonomous vehicles with kinematic constraints. Piecewise linear continuous-curvature paths composed of clothoids, circular arcs, and straight lines are used for this purpose. Low curvature derivatives provide comfort to passengers. This approach provides a compact road network representation as only the parameters of the curves are stored. A real urban scenario with straight and curved roads, multiple lanes, intersections, and roundabouts is modeled and the proposed approach is validated. As a result of the proposed approaches, door-to-door global continuous-curvature paths are generated considering the shortest traveled distance.

In general, it is not possible to determine the actual manipulator platform's pose of a parallel mechanism from its active joints' coordinates. This problem is usually solved by using additional numerical procedures or by additional system information from auxiliary sensors, providing several weaknesses including initial pose estimations, reference drives, or workspace limitations. In this paper, we therefore introduce a closed-form solution for the direct kinematics problem of the planar 3-RPR parallel mechanism by using only the orientations of two active joints and the manipulator platform, where P denotes active prismatic joints and R passive revolute joints.

For an autonomous mobile robot, when the available power goes below a certain threshold, the robot needs to abort its current task and move towards a charging station to recharge its battery. The efficiency of an autonomous mobile robot depends significantly on the location of the charging stations. In this paper, we address the charging station placement problem for mobile robots in a controlled workspace. We propose two algorithms to place a number of charging stations so that a robot is always capable of reaching one of the charging stations from any obstacle-free location in the workspace without aborting its task too early. We reduce the charging-station placement problem to a series of Satisfiability Modulo Theory (SMT) problems and use the off-the-shelf SMT solver Z3 to implement our algorithm. The algorithm produces as output the locations of the charging stations in the workspace and the trajectories from any obstacle-free locations to one of the charging stations. Our experimental results show how our algorithm can efficiently find the locations of the charging stations and robot trajectories to reach the charging stations. We demonstrate through simulation how the generated trajectories can be effectively used by a robot to reach a charging stations autonomously without getting depleted with power.

Inspired by the locomotor nervous system of vertebrates, central pattern generator (CPG) models can be used to design gaits for articulated robots, such as crawling, swimming or legged robots. Incorporating sensory feedback for gait adaptation in these models can improve the locomotive performance of such robots in challenging terrain. However, many CPG models to date have been developed exclusively for open-loop gait generation for traversing level terrain. In this paper, we present a novel approach for incorporating inertial feedback into the CPG framework for the control of body posture during legged locomotion on steep, unstructured terrain. That is, we adapt the limit cycle of each leg of the robot with time to simultaneously produce locomotion and body posture control. We experimentally validate our approach on a hexapod robot, locomoting in a variety of steep, challenging terrains (grass, rocky slide, stairs). We show how our approach can be used to level the robot's body, allowing it to locomote at a relatively constant speed, even as terrain steepness and complexity prevents the use of an open-loop control strategy.

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.

The existence of tactile afferents sensitive to slip-related mechanical transients in the human hand augments the robustness of grasping through secondary force modulation protocols. Despite this knowledge and the fact that tactile-based slip detection has been researched for decades, robust slip detection is still not an out-of-the-box capability for any commercially available tactile sensor. This research seeks to bridge this gap with a comprehensive study addressing several aspects of slip detection. In particular, key developments include a systematic data collection process yielding millions of sensory data points, a spectral analysis of sensory responses providing insight into sensor behavior, and the application of Long Short-Term Memory (LSTM) neural networks to produce robust slip detectors from three commercially available sensors capable of tactile sensing. The sensing mechanics behind these sensors are all fundamentally different and leverage principles in electro-mechanical resistance, optics, and hydro-acoustics. Critically, slip detection performance of the tactile technologies is quantified through a measurement methodology that unveils the effects of data window size, sampling rate, material type, slip speed, and sensor manufacturing variability. Results indicate that the investigated commercial tactile sensors are inherently capable of high-quality slip detection.

We propose to synthesize feasible caging grasps for a target object through computing Caging Loops, a closed curve defined in the shape embedding space of the object. Different from the traditional methods, our approach decouples caging loops from the surface geometry of target objects through working in the embedding space. This enables us to synthesize caging loops encompassing multiple topological holes, instead of always tied with one specific handle which could be too small to be graspable by the robot gripper. Our method extracts caging loops through a topological analysis of the distance field defined for the target surface in the embedding space, based on a rigorous theoretical study on the relation between caging loops and the field topology. Due to the decoupling, our method can tolerate incomplete and noisy surface geometry of an unknown target object captured on-the-fly. We implemented our method with a robotic gripper and demonstrate through extensive experiments that our method can synthesize reliable grasps for objects with complex surface geometry and topology and in various scales.

Localization is an essential component for autonomous robots. A well-established localization approach combines ray casting with a particle filter, leading to a computationally expensive algorithm that is difficult to run on resource-constrained mobile robots. We present a novel data structure called the Compressed Directional Distance Transform for accelerating ray casting in two dimensional occupancy grid maps. Our approach allows online map updates, and near constant time ray casting performance for a fixed size map, in contrast with other methods exhibit poor worst case performance. Our experimental results show that the proposed algorithm approximates the performance characteristics of reading from a three dimensional lookup table of ray cast solutions while requiring two orders of magnitude less memory and precomputation. This results in a particle filter algorithm which can maintain 2500 particles with 61 ray casts per particle at 40Hz, using a single CPU thread onboard a mobile robot.

Recent self-supervised learning approaches focus on using a few thousand data points to learn policies for high-level, low-dimensional action spaces. However, scaling this framework for higher-dimensional control requires either scaling up the data collection efforts or using a clever sampling strategy for training. We present a novel approach - Curriculum Accelerated Self-Supervised Learning (CASSL) - to train policies that map visual information to high-level, higher-dimensional action spaces. CASSL orders the sampling of training data based on control dimensions: the learning and sampling are focused on few control parameters before other parameters. The right curriculum for learning is suggested by variance-based global sensitivity analysis of the control space. We apply our CASSL framework to learning how to grasp using an adaptive, underactuated multi-fingered gripper, a challenging system to control. Our experimental results indicate that CASSL provides significant improvement and generalization compared to baseline methods such as staged curriculum learning (8% increase) and complete end-to-end learning with random exploration (14% improvement) tested on a set of novel objects.

This paper shows and evaluates a novel approach to integrate a non-invasive Brain-Computer Interface (BCI) with the Robot Operating System (ROS) to mentally drive a telepresence robot. Controlling a mobile device by using human brain signals might improve the quality of life of people suffering from severe physical disabilities or elderly people who cannot move anymore. Thus, the BCI user can actively interact with relatives and friends located in different rooms thanks to a video streaming connection to the robot. To facilitate the control of the robot via BCI, we explore new ROS-based algorithms for navigation and obstacle avoidance in order to make the system safer and more reliable. In this regard, the robot exploits two maps of the environment, one for localization and one for navigation, and both are used as additional visual feedback for the BCI user to control the robot position. Experimental results show a decrease of the number of commands needed to complete the navigation task, suggesting a reduction user's cognitive workload. The novelty of this work is to provide a first evidence of an integration between BCI and ROS that can simplify and foster the development of software for BCI driven robotics devices.

Handling loop closures and intervehicle observations in cooperative robotic scenarios remains a challenging problem due to data consistency, bandwidth limitations and increased computation requirements. This paper develops a general cooperative localisation and single vehicle Visual SLAM framework that includes direct intervehicle observations and pose to pose loop closures on each vehicle with states shared as required. This fuses single vehicle SLAM with cooperative localisation and avoids data association of map data across limited communication networks. The base problem is developed as a factor graph with each vehicle solving local subgraphs that are split based on intervehicle observations. We modify the order of variable elimination in subgraphs through manipulation of the square-root of the Information matrix to extract updates that include the historic states involved in the loop closures and do not require transmission of other states not involved in the measurement or retransmission of previously shared states. We demonstrate the effect on localisation accuracy and bandwidth using data captured from a set of five robots observing each other and landmarks compared to both single vehicle SLAM, pure cooperative localisation and a centralised solution.

Being aware of our body has great importance in our everyday life. It helps us to complete difficult tasks, such as movement in a dark room or grasping a complex object. These skills are important for robots as well, however, robotic bodily awareness is still an open question, and the nonlinearity of soft robots adds even more complexity. In this paper, we address this problem and present a novel method to implement bodily awareness into a real soft robot by the integration of its exteroceptive and proprioceptive sensors. We use an octopus-inspired arm as an example where the proprioceptive representation is approximated by four bend sensors integrated into the soft body, while a camera records the movement of the arm capturing its exteroceptive representation. The internal sensory signals are mapped to the visual information using a combination of a stacked convolutional autoencoder (CAE) and a recurrent neural network (RNN). As a result, the soft robot can learn to estimate and, therefore, to imagine its motion even when its visual sensor is not available.

Most robotics literature model the human's knee and hip as a revolute joint with limited range of rotation. Although somehow close to reality, this approach neglects a critical aspect of these joints, which is their internal flexibility. This paper presents a prototype tensegrity flexural manipulator whose kinematic behavior is inspired by human leg's gait. This prototype, which considers a hybrid (flexible-rigid) structure of the knee and hip would be able to better approximate real behavior and hopefully lead to a better design of artificial (prosthetic) knees and hips. The behavior of the proposed tensegrity manipulator was firstly predicted using OpenSim simulation environment. The paper reports the comparisons between the simulations, physical prototypes and human leg behavior for a variety of ranges of motions and tension analysis.

Many modern roboticists take inspiration from biology to create novel robotic structures, including those that are modeled after the octopus. This paper advances this trend by creating soft robots modeling the complex motion patterns of octopus tentacles employing a bio-mimetic approach. The proposed octopus robot is entirely made from soft material and uses a novel fluidic actuation mechanism that allows the robot to advance forward, change directions and rotate around its primary axis. The paper presents the robot's design and fabrication process. An experimental study is conducted showing the feasibility of the proposed robot and actuation mechanism.

This paper introduces geometry and object shape and pose costs for multi-object tracking in urban driving scenarios. Using images from a monocular camera alone, we devise pairwise costs for object tracks, based on several 3D cues such as object pose, shape, and motion. The proposed costs are agnostic to the data association method and can be incorporated into any optimization framework to output the pairwise data associations. These costs are easy to implement, can be computed in real-time, and complement each other to account for possible errors in a tracking-by-detection framework. We perform an extensive analysis of the designed costs and empirically demonstrate consistent improvement over the state-of-the-art under varying conditions that employ a range of object detectors, exhibit a variety in camera and object motions, and, more importantly, are not reliant on the choice of the association framework. We also show that, by using the simplest of associations frameworks (two-frame Hungarian assignment), we surpass the state-of-the-art in multi-object-tracking on road scenes. More qualitative and quantitative results can be found at https://junaidcs032.github.io/Geometry_ObjectShape_MOT/.

We introduce an algorithm for autonomous control of multiple fast ground vehicles operating in close proximity to each other. The algorithm is based on a combination of the game theoretic notion of iterated best response, and an information theoretic model predictive control algorithm designed for non-linear stochastic systems. We test the algorithm on two one-fifth scale AutoRally platforms traveling at speeds upwards of 8 meters per second, while maintaining a following distance of under two meters from bumper-to-bumper.

We propose an algorithm for robust visual classification of an object of interest observed from multiple views using a black-box Bayesian classifier which provides a measure of uncertainty, in the presence of significant ambiguity and classifier noise, and of localization error. The fusion of classifier outputs takes into account viewpoint dependency and spatial correlation among observations, as well as pose uncertainty when these observations are taken and a measure of confidence provided by the classifier itself. Our experiments confirm an improvement in robustness over state-of-the-art.

We propose a novel real-time algorithm for estimating the local scale correction of a monocular SLAM system, to obtain a correctly scaled version of the 3D map and of the camera trajectory. Within a Bayesian framework, it integrates observations from a deep-learning based generic object detector and landmarks from the map whose projection lie inside a detection region, to produce scale correction estimates from single frames. For each observation, a prior distribution on the height of the detected object class is used to define the observation's likelihood. Due to the scale drift inherent to monocular SLAM systems, we also incorporate a rough model on the dynamics of scale drift. Quantitative evaluations are presented on the KITTI dataset, and compared with different approaches. The results show a superior performance of our proposal in terms of relative translational error when compared to other monocular systems based on object detection.