ICRA 2015 Papers

Binaural sound source localization is an important technique involving Human-Robot Interaction (HRI), video conference, speech enhancement, etc. In many real application scenarios, especially for closed environments, the affect of reverberation and noise would degrade the precision of position estimations. Therefore, a new binaural sound source localization method based on generalized parametric model and two-layer matching strategy is proposed in this paper for complex environments. Firstly, cepstral prefiltering is utilized for dereverberation of binaural signals. Then, two binaural cues computed from a dual-channel frequency representation, are combined to estimate the azimuths of sources. Additionally, the generalized parametric model is presented to describe the relationship between the azimuth and binaural cues through finding the optimal scaling factors from training data. At last, a two-layer matching strategy based on Bayesian rule is used to make the final decision, which can effectively decrease the computation complexity. Experiments have validated the proposed approach and show that it achieves favorably better results compared with several available methods without extra spacial burden.

Given that 3D scan matching is such a central part of the perception pipeline for robots, thorough and large-scale investigations of scan matching performance are still surprisingly few. A crucial part of the scientific method is to perform experiments that can be replicated by other researchers in order to compare different results. In light of this fact, this paper presents a thorough comparison of 3D scan registration algorithms using a recently published benchmark protocol which makes use of a publicly available challenging data set that covers a wide range of environments. In particular, we evaluate two types of recent 3D registration algorithms - one local and one global. Both approaches take local surface structure into account, rather than matching individual points. After well over 100 000 individual tests, we conclude that algorithms using the normal distributions transform (NDT) provides accurate results compared to a modern implementation of the iterative closest point (ICP) method, when faced with scan data that has little overlap and weak geometric structure. We also demonstrate that the minimally uncertain maximum consensus (MUMC) algorithm provides accurate results in structured environments without needing an initial guess, and that it provides useful measures to detect whether it has succeeded or not. We also propose two amendments to the experimental protocol, in order to provide more valuable results in future implementations.

We consider the problem of learning from demonstrations to manipulate deformable objects. Recent work [1], [2], [3] has shown promising results that enable robotic manipulation of deformable objects through learning from demonstrations. Their approach is able to generalize from a single demonstration to new test situations, and suggests a nearest neighbor approach to select a demonstration to adapt to a given test situation. Such a nearest neighbor approach, however, ignores important aspects of the problem: brittleness (versus robustness) of demonstrations when generalized through this process, and the extent to which a demonstration makes progress towards a goal. In this paper, we frame the problem of selecting which demonstration to transfer as an options Markov decision process (MDP). We present max-margin Q-function estimation: an approach to learn a Q-function from expert demonstrations. Our learned policies account for variability in robustness of demonstrations and the sequential nature of our tasks. We developed two knot-tying benchmarks to experimentally validate the effectiveness of our proposed approach. The selection strategy described in [2] achieves success rates of 70% and 54%, respectively. Our approach performs significantly better, with success rates of 88% and 76%, respectively.

We present a Bayesian approach to tactile object recognition that improves on state-of-the-art in using single-touch events in two ways. First by improving recognition accuracy from about 90% to about 95%, using about half the number of touches. Second by reducing the number of touches needed for training from about 200 to about 60. In addition, we use a new tactile sensor that is less than one tenth of the cost of widely available sensors. The paper describes the sensor, the likelihood function used with the Naive Bayes classifier, and experiments on a set of ten real objects. We also provide preliminary results to test our approach for its ability to generalise to previously unencountered objects.

Typically 2.5D maps provide a compact and efficient representation of the environment. When sensor data is obtained from multiple sets of noisy measurements at differing resolutions, the problem of compounding this information together to provide an effective and efficient means of mapping is not trivial, particularly as the size of the environment increases. In this paper, we propose a general framework for integrating heterogeneous sensor data to obtain large-scale 2.5D probabilistic maps. Gaussian Processes are used to generate a prior map that learns the spatial correlation between nearby points. Bayesian data fusion is then employed to update these prior maps with new measurements from distinct sensor modalities. In order to deal with large scale data, a novel submapping strategy is introduced to perform the fusion step efficiently in dealing with large covariance matrices. Submaps are first marginalised from the learned correlated prior and then updated based on the property of conditional independence. Most notably, the technique lends itself to generate accurate estimates at arbitrary resolutions and is able to handle varying noise from disparate sensor sources. The framework is applied to pipeline thickness mapping, with experimental results in fusing a high-resolution sensor and a low-resolution sensor showing the ability of the proposed technique to capture spatial correlations to come up with more accurate results when compared with a naïve fusion approach.

We propose and experimentally validate a new class of models for the total thrust generation in multi-rotor UAVs which is suitable for low- and middle-end platforms. Differently from typical models assuming to instantaneously control the rotor spinning velocity, in the proposed class we consider that the total thrust has its own dynamics and its final value explicitly depends both on the pseudo-setpoint commands given to the motor driver and the measurement of the battery terminal voltage. We compare the different model instances within the class using a principled experimental setup in which the total thrust is precisely measured using a motion capture system as ground truth, instead of relying on a setup based or noise-prone force sensors. We then show that the use of a dynamical model that includes also the battery terminal voltage significantly improves the prediction ability of the model in terms of accuracy. Finally we show how the proposed model can be identified using on-board only acceleration measurements, achieving a surprisingly good accuracy when compared with the ground truth case. We expect that the use of the proposed model will be important both in case of precise flight control and in the case of aerial physical interactive tasks.

In this paper, we present Batch Informed Trees (BIT*), a planning algorithm based on unifying graph- and sampling-based planning techniques. By recognizing that a set of samples describes an implicit random geometric graph (RGG), we are able to combine the efficient ordered nature of graph-based techniques, such as A*, with the anytime scalability of sampling-based algorithms, such as Rapidly-exploring Random Trees (RRT).

This paper proposes a momentum-based balancing controller for robots which have non-rigid contacts with their environments. This controller regulates both linear momentum and angular momentum about the center of mass of the robot by controlling the contact forces. Compliant contact models are used to determine the contact forces at the contact points. Simulation results show the performance of the controller on a four-link planar robot standing on various compliant surfaces while unknown external forces in different directions are acting on the center of mass of the robot.

We present the design of a backstepping controller for a triangular over-actuated marine platform, controlled by three rotating jets. Our goal is the stabilization of the position and the orientation of the platform, under realistic environmental disturbances, such as wind forces, wave forces and hydrodynamic forces. Actuator thrust and angle dynamics, as well as settling delays, in the rotation of the jets and in the response of the desired thrust, are included in the algorithm, despite the presence of an allocation scheme. Thrust and angle velocity, limitations are also taken into account. A Thrust Upper Limit (TUL) manipulation heuristic is introduced in order to reduce the thrust requirements and the energy consumed. The performance of the developed backstepping controller is compared to the case with and without the TUL. Simulation results show that the use of the heuristic reduces energy consumption.

This paper proposes a controller for the GIRONA500 Underwater Vehicle Manipulator System and reports experimental results for an autonomous floating-base valve-turning manipulation application. The selected method is based on kinematic control, avoiding the need for a complex, and difficult to obtain, hydrodynamic model. The method relies on the decoupled control of the vehicle and manipulator velocities using a combination of the task priority redundancy resolution and the task concurrence approaches. The paper discusses the manipulation tasks needed, their hierarchical organisation, and a set of strategies that were needed to complement the `standard' task-priority approach to successfully solve the floating-base manipulation problem in water tank experimental conditions.

Gliding robotic fish, a hybrid of underwater gliders and robotic fish, are energy-efficient and highly maneuverable, and hold strong promise for long-duration sampling of underwater environments. In this paper a novel systematic autonomous water-column-based sampling scheme for gliding robotic fish is proposed to measure the three-dimensional spatial distributions of variables of interest in aquatic environments. The scheme exploits energy-efficient spiral-down motion to sample each water column, followed by sagittal-plane glide-up towards the direction of next water column. Once surfacing, the robot uses GPS guidance to reach the next column location through swimming. To enhance the path tracking performance, a two-degree-of-freedom controller involving H∞ control is used in the spiral motion, and a sliding-mode controller is employed to regulate the yaw angle during glide-up. The sampling scheme has been implemented on a gliding robotic fish prototype, “Grace”, and verified first in pool experiments, and then in field experiments involving the sampling of harmful algae concentration in the Wintergreen Lake, Michigan.

In this paper we examine the problem of autonomously exploring and mapping an environment using a mobile robot. The robot uses a graph-based SLAM system to perform mapping and represents the map as an occupancy grid. In this setting, the robot must trade-off between exploring new area to complete the task and exploiting the existing information to maintain good localization. Selecting actions that decrease the map uncertainty while not significantly increasing the robot's localization uncertainty is challenging. We present a novel information-theoretic utility function that uses both Shannon's and Rényi's definitions of entropy to jointly consider the uncertainty of the robot and the map. This allows us to fuse both uncertainties without the use of manual tuning. We present simulations and experiments comparing the proposed utility function to state-of-the-art utility functions, which only use Shannon's entropy. We show that by using the proposed utility function, the robot and map uncertainties are smaller than using other existing methods.

This paper details an autonomous perching and take-off method for a quadrotor micro air vehicle (MAV) using a novel dry adhesive gripper on smooth vertical walls. The gripper mechanism uses three directional dry adhesive pads in a triangular configuration. Each pad is equipped with a force sensor that can detect the pad's loading condition. A servo motor is used to actuate the attachment and detachment of the gripper, which is mounted in the front of a quadrotor MAV. This makes perching possible by simply flying toward and hitting the target surface. Autonomous control is made possible using a Microsoft Kinect to localize the MAV and a PID controller to control the perching maneuver. Experiments show that a minimum speed of 0.4m/s is required to guarantee a successful perch. Also, in 93% of the experiments in which the MAV hits the target at a speed higher than 0.4m/s, the perching maneuver is successful. To initiate a take-off procedure, a release signal is sent to the servo and the gripper is detached from the wall by pulling the adhesive away from the surface. Once the gripper is detached, the MAV becomes airborne again and the control system stabilizes the flight.

Autonomous navigation of generic monocular quadcopter in the natural environment requires sophisticated mechanism for perception, planning and control. In this work, we have described a framework which performs perception using monocular camera and generates minimum time collision free trajectory and control for any commercial quadcopter flying through cluttered unknown environment. The proposed framework first utilizes supervised learning approach to estimate the dense depth map for video stream obtained from frontal monocular camera. This depth map is initially transformed into Ego Dynamic Space and subsequently, is used for computing locally traversable way-points utilizing binary integer programming methodology. Finally, trajectory planning and control module employs a convex programming technique to generate collision-free trajectory which follows these way-points and produces appropriate control inputs for the quadcopter. These control inputs are computed from the generated trajectory in each update. Hence, they are applicable to achieve closed-loop control similar to model predictive controller. We have demonstrated the applicability of our system in controlled indoors and in unstructured natural outdoors environment.

This paper proposes an autonomous landing scheme for a small-scale flybarless helicopter equipped with low-cost navigation sensors. The main contribution of this paper is the design of a model-free motion controller that guarantees autonomous landing with prescribed transient and steady state response, despite the presence of external disturbances acting on the vehicle. The proposed control scheme is of low complexity and does not require any knowledge of the helicopter dynamic parameters. Hence, it can be easily implemented in embedded control platforms integrated on small-scale helicopters. In order to provide the controller with accurate estimation of the vehicle's state vector during the landing procedure, an asynchronous sensor fusion and state estimation algorithm, based on an Unscented Kalman Filter (UKF), has been also implemented. The performance and the efficiency of the overall scheme are experimentally verified using a small-scale flybarless helicopter in a real autonomous landing process.

In this paper, we investigate the question of how a legged robot can walk efficiently by taking advantage of its ability to alter its gait as a function of statistical (large-scale) terrain properties. One of the contributions of this paper is the algorithm to achieve real-time terrain identification and autonomous gait adaptation on a legged robot. We approach this problem by first classifying the terrains based on their proprioceptive responses and identifying the terrain in real-time. Then we choose an optimal gait to best suit the identified terrain type. We exploit our recent findings regarding gaits, estimated from terrain-contact signatures, in order to obtain an optimized mapping between terrain signatures and terrain-specific gaits. We evaluate our algorithm on synthetic data, and real robot data collected on different terrains and naturally occurring terrain transitions. Another key contribution of this work is the statistical verification that precise gait selection can lead to energy savings in practice in legged robots. This assessment of energy efficiency, achieved by gait adaptation, is among the firsts of its kind in gait adaptation literature. We also present an analysis of the effect of terrain transition frequency on our gait adaptation algorithm. Our results are supported by validation using both synthetic data and field testing.

Survey class Autonomous Underwater Vehicles (AUVs) typically rely on Doppler Velocity Logs (DVL) for precise navigation near the seafloor. In cases where the distance to the seafloor is greater than the DVL bottom lock range, localizing between the surface where GPS is available and the seafloor presents a localization problem, since both GPS and DVL are unavailable in the mid-water column. Previous work proposed a solution to navigation in the mid-water column that exploits the stability of the vertical water current profile in space over the minutes scale. With repeated measurements of these currents with the Acoustic Doppler Current Profiler (ADCP) mode of the DVL during vertical descent, along with sensor fusion of other low cost sensors, position error growth is constrained to near the initial velocity uncertainty of the vehicle at the sea surface during the dive. Following DVL bottom lock, due to correlations in the joint vehicle and water current velocity estimation, the entire velocity history is further constrained. In this paper a number of extensions are developed to improve navigation performance during missions characterized by prolonged time-scales and horizontal transits. This includes explicitly accounting for the temporal evolution of water currents while revisiting locations and during the re-observation itself, in addition to employing high fidelity spatial models to account for the horizontal water current field. Furthermore, the validity of our method is demonstrated on deep water data sets. Results from the Sentry AUV are presented from a hydrothermal vent flux estimation mission and a surveying mission at Deepwater Horizon, illustrating a real-time dead-reckoning method with results for up to 20 hours.

Autonomous, vision-based quadrotor flight is widely regarded as a challenging perception and control problem since the accuracy of a flight maneuver is strongly influenced by the quality of the on-board state estimate. In addition, any vision-based state estimator can fail due to the lack of visual information in the scene or due to the loss of feature tracking after an aggressive maneuver. When this happens, the robot should automatically re-initialize the state estimate to maintain its autonomy and, thus, guarantee the safety for itself and the environment. In this paper, we present a system that enables a monocular-vision-based quadrotor to automatically recover from any unknown, initial attitude with significant velocity, such as after loss of visual tracking due to an aggressive maneuver. The recovery procedure consists of multiple stages, in which the quadrotor, first, stabilizes its attitude and altitude, then, re-initializes its visual state-estimation pipeline before stabilizing fully autonomously. To experimentally demonstrate the performance of our system, we aggressively throw the quadrotor in the air by hand and have it recover and stabilize all by itself. We chose this example as it simulates conditions similar to failure recovery during aggressive flight. Our system was able to recover successfully in several hundred throws in both indoor and outdoor environments.

In many robotic processes such as milling and drilling, there are multiple solutions for robot poses, as the rotation around the tool axis remains as a degree of freedom (DoF) in positioning. Yet until now, this DoF causes additional efforts in CAM programming as it requires manual intervention. Instead, this DoF can be used to optimize the robot pose according to different criteria of the robot such as stiffness or avoidance of backlash effects. This paper presents different criteria for optimization of the robot pose in machining and describes the optimization of robot stiffness based on a novel method for its identification, which is in detail described on this paper. Furthermore, the potential of the automatic resolution of the DoF is outlined enabling staff without robot knowledge to define reasonable robot paths.