ICRA 2015 Papers

This paper reports on the development and validation of the elastodynamics model of an innovative underwater soft-bodied robot inspired by cephalopods. The vehicle, for which the model is devised, is propelled by a discontinuous activation routine which entails the collapse of an elastic shell via cable transmission and its following passive re-inflation under the action of the elastic energy stored in the shell walls. Activation routine and thrust characterization have been determined to depend massively on the capability of the shell to elastically return to its unstrained state, hence an accurate description of the dynamics of the shell during all stages of actuation and at various degrees of deformation is essential. The model, based on a geometrically exact Cosserat theory, is validated against measurement achieved from an ad-hoc experimental apparatus, bringing evidence of its aptness at capturing the key parameters of the system. Eventually the model is employed for simulating a proper propulsion routine in water demonstrating that, upon suitable parametrization of the internal and external hydrodynamics, it can reliably be employed for the realistic quantitative characterization of the cephalopod-inspired robot.

Maps used for navigation often include a database of location descriptions for place recognition (loop closing), which permits bounded-error performance. A standard pose-graph SLAM system adds a new entry for every new pose into the location database, which grows linearly and unbounded in time and thus becomes unsustainable. To address this issue, in this paper we propose a new map-reduction approach that pre-constructs a fixed-size place-recognition database amenable to the limited storage and processing resources of the vehicle by exploiting the high-level structure of the environment as well as the vehicle motion. In particular, we introduce the concept of location utility - which encapsulates the visitation probability of a location and its spatial distribution relative to nearby locations in the database - as a measure of the value of potential loop-closure events to occur at that location. While finding the optimal reduced location database is NP-hard, we develop an efficient greedy algorithm to sort all the locations in a map based on their relative utility without access to sensor measurements or the vehicle trajectory. This enables pre-determination of a generic, limited-size place-recognition database containing the N best locations in the environment. To validate the proposed approach, we develop an open-source street-map simulator using real city-map data and show that an accurate map (pose-graph) can be attained even when using a place-recognition database with only 1% of the entries of the corresponding full database.

With the growing demand for deployment of robots in real scenarios, robustness in the perception capabilities for navigation lies at the forefront of research interest, as this forms the backbone of robotic autonomy. Existing place recognition approaches traditionally follow the feature-based bag-of-words paradigm in order to cut down on the richness of information in images. As structural information is typically ignored, such methods suffer from perceptual aliasing and reduced recall, due to the ambiguity of observations. In a bid to boost the robustness of appearance-based place recognition, we consider the world as a continuous constellation of visual words, while keeping track of their covisibility in a graph structure. Locations are queried based on their appearance, and modelled by their corresponding cluster of landmarks from the global covisibility graph, which retains important relational information about landmarks. Complexity is reduced by comparing locations by their graphs of visual words in a simplified manner. Test results show increased recall performance and robustness to noisy observations, compared to state-of-the-art methods.

This paper presents a novel framework for integration of vision and tactile sensing by localizing tactile readings in a visual object map. Intuitively, there are some correspondences, e.g., prominent features, between visual and tactile object identification. To apply it in robotics, we propose to localize tactile readings in visual images by sharing same sets of feature descriptors through two sensing modalities. It is then treated as a probabilistic estimation problem solved in a framework of recursive Bayesian filtering. Feature-based measurement model and Gaussian based motion model are thus built. In our tests, a tactile array sensor is utilized to generate tactile images during interaction with objects and the results have proven the feasibility of our proposed framework.

To determine the pose of a vehicle is a fundamental problem in mobile robotics. Most approaches relate the current sensor observations to a map generated with previously acquired data of the same system or by another system with a similar sensor setup. Unfortunately, previously acquired data is not always available. In outdoor settings, GPS is a very useful tool to determine a global estimate of the vehicles pose. Unfortunately, GPS tends to be unreliable in situations in which a clear view to the sky is restricted. Yet, one can make use of publicly available map material as prior information. In this paper, we describe an approach to localize a robot equipped with a 3D range scanner with respect to a road network created from OpenStreetMap data. To successfully localize a mobile robot we propose a road classification scheme for 3D range data together with a novel sensor model, which relates the classification results to a road network. Compared to other approaches, our system does not require the robot to actually travel on the road network. We evaluate our approach in extensive experiments on simulated and real data and compare favorably to two state-of-the-art methods on those data.

We present and analyze methods for patrolling and surveillance in an environment with a distributed swarm of robots with limited capabilities. Our approach is based on a distributed triangulation of the work space, in which a set of p stationary sensors provides coverage control; in addition, there are r mobile robots that can move between the sensors. Building on our prior work on structured exploration of unknown spaces with multi-robot systems, we can make use of a triangulation that is constructed in a distributed fashion and guarantees good local navigation properties, even when sensors and robots have very limited capabilities.

In this work, we deal with the problem of planning a manipulation task for a robotic system composed of at least one dexterous arm and a dexterous multi-fingered hand. The goal of the local planner is to include both, the arm and the hand, in the execution of the task in a coordinated way. This is achieved by using the workspace of the hand which is computed offline. During the online planning, the current in-hand manipulation capability is evaluated taking advantage of the dimensions of the hand workspace and considering the task itself. Dynamic weights enable the computation of the instantaneous contributions of the two subsystems on the motion of the manipulated object. The method is evaluated in simulation as well as in several experiments on the real robot.

This paper presents a formalism that exploits deformability during manipulation of soft objects by robot teams. A hybrid centralized/distributed approach restricts centralized planning to high-level global guidance of the object for consensus. Low-level control is thus delegated to the individual manipulator robots, which retain manipulation and collision avoidance guarantees by passing forces to one another through the object. A distributed receding horizon planner provides local control, formulated as a convex optimization problem in velocity space and incorporating constraints for both collision avoidance and shape maintenance. We demonstrate teams of mobile manipulators autonomously carrying various deformable objects.

Our research group has started a collaboration that analyzes data collected from biological snakes to provide insight on how to better program snake robots. Most data collected on biological snakes views the snakes from above and thus can only detect motion in the horizontal plane. However, both our robots and biological snakes are capable of generating motions both in the horizontal and vertical planes. Vertical waves naturally play a major role in limbless locomotion in that they simultaneously provide thrust motion and make-and-break contact between the mechanism and environment. Analysis on the data, collected from sidewinder rattle snakes, revealed that disparate modes of locomotion emerged from different contact patterns. We conclude that the same horizontal undulation can cause dramatically different motions for both the biological and robotic snakes depending upon the choice of contacts. With this knowledge, we introduce contact scheduling, a technique that plans positions of contacts along the body to design gaits for snake robots. Contact scheduling results in a novel turning gait, which can reorient a snake robot more than 90 degrees in one gait cycle.

Fluid-mediated programmable stochastic self-assembly offers promising means to formation of target structures capable of a variety of functionalities. While miniaturized building blocks allow for finer resolutions in such structures, as well as access to unconventional environments, they can only be endowed with very limited on-board resources. In this paper we present the design, fabrication, and experimental results validating the key functionalities of the Lily robot as the building block in a programmable stochastic fluidic self-assembly system, capable of forming 2D structures. In particular, we aim at driving a system including an arbitrary number of Lilies to form target structures through parallel self-assembly, using exclusively local information and communication. While capable of wireless communication to a base station, Lilies are endowed with custom-designed electropermanent magnets to latch and also to communicate locally with their neighbors. Several experiments validate the reliability of the radio channel as well as the robustness of the local induction-based communication which allows for data transfer at 9600 bps with a success rate of 92.8% without repetition. The latches are shown to hold four times the weight of a single robot and to drag in another Lily from a distance of 4 mm in water.

We propose an algorithm to estimate the “lifetime” of events from retinal cameras, such as a Dynamic Vision Sensor (DVS). Unlike standard CMOS cameras, a DVS only transmits pixel-level brightness changes (“events”) at the time they occur with micro-second resolution. Due to its low latency and sparse output, this sensor is very promising for high-speed mobile robotic applications. We develop an algorithm that augments each event with its lifetime, which is computed from the event's velocity on the image plane. The generated stream of augmented events gives a continuous representation of events in time, hence enabling the design of new algorithms that outperform those based on the accumulation of events over fixed, artificially-chosen time intervals. A direct application of this augmented stream is the construction of sharp gradient (edge-like) images at any time instant. We successfully demonstrate our method in different scenarios, including high-speed quadrotor flips, and compare it to standard visualization methods.

This paper describes the application of a weight estimation method for emergency patients in clinical environments. The approach applies established algorithms for point cloud processing and filtering to data from a low-cost, structured light sensor. A patient's volume is estimated on the basis of their visible front surface. The approach is currently being tested in the workflow of the emergency room at the Universitätsklinikum Erlangen, Germany. Preliminary results show the accuracy of the approach in relation to other conservative means of weight measurements, for example, by physicians and anthropometric measurements.

In the last years, many researchers started to consider teach-and-repeat approaches for reliable autonomous navigation. The paradigm, in all its proposed forms, is deeply rooted in the idea that the robot should autonomously follow a route that has been demonstrated by a human during a teach phase. However, human demonstrations are often inefficient in terms of execution time or may cause premature wear of the robot components due to jittery behavior or strong accelerations. In this paper, we propose the concept of teach, optimize and repeat, which introduces a trajectory optimization step between the teach and repeat phases. To address this problem, we further propose LexTOR, a constrained trajectory optimization method for teach and repeat problems, where the constraints are defined according to user preferences. At its core, LexTOR optimizes both the execution time and the trajectory smoothness in a lexicographic sense. The experiments show that LexTOR is very effective, both qualitatively and quantitatively, in terms of execution time, smoothness, accuracy and bound satisfaction.

In this paper, we address the challenging problem of single-view cross-season place recognition. A new approach is proposed for compact discriminative scene descriptor that helps in coping with changes in appearance in the environment. We focus on a simple effective strategy that uses objects whose appearance remain the same across seasons as valid landmarks. Unlike popular bag-of-words (BoW) scene descriptors that rely on a library of vector quantized visual features, our descriptor is based on a library of raw image data (e.g., visual experience shared by colleague robots, publicly available photo collections from Google StreetView), and directly mines it to identify landmarks (i.e., image patches) that effectively explain an input query/database image. The discovered landmarks are then compactly described by their pose and shape (i.e., library image ID, and bounding boxes) and used as a compact discriminative scene descriptor for the input image. We collected a dataset of single-view images across seasons with annotated ground truth, and evaluated the effectiveness of our scene description framework by comparing its performance to that of previous BoW approaches, and by applying an advanced Naive Bayes Nearest neighbor (NBNN) image-to-class distance measure.

Recent successful approaches to autonomous vehicle localisation and navigation typically involve 3D LIDAR scanners and a static, curated 3D map, both of which are expensive to acquire and maintain. In this paper we propose an experience-based approach to matching a local 3D swathe built using a push-broom 2D LIDAR to a number of prior 3D maps, each of which has been collected during normal driving in different conditions. Local swathes are converted to a combined 2D height and reflectance representation, and we exploit the GPU rendering pipeline to densely sample the localisation cost function to provide robustness and a wide basin of convergence. Prior maps are incrementally built into an experience-based framework from multiple traversals of the same environment, capturing changes in environment structure and appearance over time. The LIDAR localisation solutions from each prior map are fused with vehicle odometry in a probabilistic framework to provide a single pose solution suitable for automated driving. Using this framework we demonstrate real-time centimetre-level localisation using LIDAR data collected in a dynamic city environment over a period of a year.

We propose a new large-scale database containing grasps that are applied to a large set of objects from numerous categories. These grasps are generated in simulation and are annotated with different grasp stability metrics. We use a descriptive and efficient representation of the local object shape at which each grasp is applied. Given this data, we present a two-fold analysis: (i) We use crowdsourcing to analyze the correlation of the metrics with grasp success as predicted by humans. The results show that the metric based on physics simulation is a more consistent predictor for grasp success than the standard υ-metric. The results also support the hypothesis that human labels are not required for good ground truth grasp data. Instead the physics-metric can be used to generate datasets in simulation that may then be used to bootstrap learning in the real world. (ii) We apply a deep learning method and show that it can better leverage the large-scale database for prediction of grasp success compared to logistic regression. Furthermore, the results suggest that labels based on the physics-metric are less noisy than those from the υ-metric and therefore lead to a better classification performance.

In this paper, we propose a novel regression method that can incorporate both positive and negative training data into a single regression framework. In detail, a leveraged kernel function for non-stationary Gaussian process regression is proposed. With this new kernel function, we can vary the correlation betwen two inputs in both positive and negative directions by adjusting leverage parameters. By using this property, the resulting leveraged non-stationary Gaussian process regression can anchor the regressor to the positive data while avoiding the negative data. We first prove the positive semi-definiteness of the leveraged kernel function using Bochner's theorem. Then, we apply the leveraged non-stationary Gaussian process regression to a real-time motion control problem. In this case, the positive data refer to what to do and the negative data indicate what not to do. The results show that the controller using both positive and negative data outperforms the controller using positive data only in terms of the collision rate given training sets of the same size.

We consider two robots operating in a common workspace where each can sense the location of the other. Each robot has its own high-level task, given as a temporal logic formula, which may include requirements regarding the other robot (e.g. the robot cannot enter a room that the other robot is currently in). This task is not initially shared with the other robot; therefore, the robots when operating, may find themselves in a conflict situation. In this paper, we develop algorithms that allow the robots to automatically, and in a provably correct manner, resolve the conflict, if possible. If the automated process fails, meaning that the robots cannot both perform their tasks in the same workspace, feedback is provided to a human, that would resolve the conflict (e.g. by changing the robot's tasks). The algorithm is demonstrated with simulated and physical robots in the lab.

Robotic devices have been used in a variety of rehabilitation protocols, including gait rehabilitation after stroke. However, robotic intervention in gait therapy has only produced moderate results compared to conventional physiotherapy. We suggest a novel approach to robotic interventions which takes advantage of inter-limb coordination mechanisms. We hypothesize the existence of a mechanism of inter-leg coordination that may remain intact after a hemiplegic stroke that may be utilized to obtain functional improvement of the impaired leg. One of the most significant advantages of this approach is the safety of the patient, since this does not require any direct manipulation of the impaired leg. In this paper, we focus on designing and applying unilateral perturbations that evoke contralateral leg motions through mechanisms of inter-leg coordination. Real-time control of floor stiffness is utilized to uniquely differentiate force and kinematic feedback, creating novel perturbations. We present results of repeatable and scalable evoked muscle activity of the contralateral tibialis anterior muscle through unilateral stiffness perturbations. We also present a mathematical model that accurately describes the relationship between the magnitude of the stiffness perturbation and the evoked muscle activity, that could result in model-based rehabilitation strategies for impaired walkers. The novel methods and results presented in this paper set the foundation for a paradigm shift of robotic interventions for gait rehabilitation.