IROS 2015 Papers

In this paper, linear regression models will be used to decode individual joint angles from low frequency EEG components. To that end, isotonic flexion/extension knee movements will be analyzed. Particularly, the decoding performance of healthy and incomplete spinal cord injured subjects will be assessed to determine the behavior of this methodology with motor disabled people. When studying cortical activity during walking, the appearance of muscular artifacts severely influences the EEG signals recorded. The analysis of single joint movements should decrease the noise provoked by the gait process itself. Additionally, different time windows prior to the decoded angle will be assessed to obtain a more reliable decoder. The results show that decoding performance is significantly above chance for most of the subjects (both healthy and disabled) and suggests that meaningful information of the movement planning starts around 2.5 seconds prior to the decoded angle.

Robotic applications in human environments are usually implemented using a cognitive architecture that integrates techniques of different levels of abstraction, ranging from artificial intelligence techniques for making decisions at a symbolic level to robotic techniques for grounding symbolic actions. In this work we address the problem of simultaneous learning at different levels of abstractions in such an architecture. This problem is important since human environments are highly variable, and many unexpected situations may arise during the execution of a task. The usual approach under this circumstance is to train each level individually to learn how to deal with the new situations. However, this approach is limited since it implies long task interruptions every time a new situation needs to be learned. We propose an architecture where learning takes place simultaneously at all the levels of abstraction. To achieve this, we devise a method that permits higher levels to guide the learning at the levels below for the correct execution of the task. The architecture is instantiated with a logic-based planner and an online planning operator learner, at the highest level, and with online reinforcement learning units that learn action policies for the grounding of the symbolic actions, at the lowest one. A human teacher is involved in the decision-making loop to facilitate learning. The framework is tested in a physically realistic simulation of the Sokoban game.

Given an articulated rigid body, we define the problem of estimating its dynamics as the problem of computing all the forces and accelerations acting on the bodies which constitute the articulated system. Similarly, we define the state estimation problem as the problem of computing the system positions and velocities. In the present paper we propose a framework for simultaneous state and dynamics estimation. The estimation is framed in a Bayesian framework and a suitable Bayesian prior is defined to guarantee the physical consistency of the obtained estimation. The Bayesian posterior makes use of all available measurements which include encoders, gyroscopes, accelerometers, force and torque sensors. The proposed theoretical framework is validated both on simulation and on the iCub humanoid. The software that implements the theoretical framework is realised with an open-source license.

The authors have developed an autonomous loading system for a wheel loader. In this paper, an algorithm is proposed to decide on both an optimized unloading point and several paths between the unloading point and several scooping points. The optimal unloading point that minimizes the total length of those paths has been obtained by using a three-dimensional configuration score space. In the configuration score space, the length of the route from origin to each grid is represented as score number, and a grid shows a position(x,y) and posture(θ). This method was evaluated through simulations and by experiments using a miniature wheel loader robot.

In this paper, an approach for sound source localisation and calibration of an asynchronous microphone array is proposed to be solved simultaneously. A graph-based Simultaneous Localisation and Mapping (SLAM) method is used for this purpose. Traditional sound source localisation using a microphone array has two main requirements. Firstly, geometrical information of microphone array is needed. Secondly, a multichannel analog-to-digital converter is required to obtain synchronous readings of the audio signal. Recent works aim at releasing these two requirements by estimating the time offset between each pair of microphones. However, it was assumed that the clock timing in each microphone sound card is exactly the same, which requires the clocks in the sound cards to be identically manufactured. A methodology is hereby proposed to calibrate an asynchronous microphone array using a graph-based optimisation method borrowed from the SLAM literature, effectively estimating the array geometry, time offset and clock difference/drift rate of each microphone together with the sound source locations. Simulation and experimental results are presented, which prove the effectiveness of the proposed methodology in achieving accurate estimates of the microphone array characteristics needed to be used on realistic settings with asynchronous sound devices.

Traditional approaches in grasping consider separately the planning of contact points and forces. This often leads to optimally simulated grasps that fail in a real execution when the forces applied at the chosen contact points cannot resist expected perturbations like the own object weight. This paper presents a method that combines a grasp planner with the computation of grasp forces, which provides contact points and contact forces that take into account joint and torque limits of the robotic hand for fingertip grasps. The planner starts with a given hand pose relative to the object, and uses the concept of reachable independent contact regions to obtain a set of force closure grasps. Within this set, the grasp that best counteracts a given external wrench is chosen. The evaluation examples illustrate how the capabilities of the robotic hand can be used more effectively by this joint planning approach.

We investigate the interaction between a pedestrian crowd and a social robot-a relevant human-robot interaction scenario observed in train stations and entrances to public places. We use a popular agent-based social force model to simulate a pedestrian crowd as it passes near a stationary robot. The modelling framework permits analysis of difficult-to-test situations such as high density crowds and the effects of contagion whereby more people are likely to interact with the robot if it is already surrounded by an audience. Accordingly, we augment the modelling framework to account for varying crowd densities, human-robot interaction, and social influence, and inform the parameter values from empirical studies in literature. Our results show that while the rate-of-interaction, defined as the number of agents located within an interacting distance per minute, increases with flow density, the average interaction-time is independent of the same. We find that inducing social influence through contagion does not have a significant effect on the rate at which agents engage with the robot in dense crowd scenarios, and has a marginal effect at low densities. The interaction-time was found to depend on the interaction-speed at which agents engage with the robot. These results indicate that at normally witnessed flow densities crowding near the robot is unlikely to take place, and the effect of robot design choices supersede the effects of social force or contagion.

We propose a novel approach for real-time object pose detection and tracking that is highly scalable in terms of the number of objects tracked and the number of cameras observing the scene. Key to this scalability is a high degree of parallelism in the algorithms employed. The method maintains a single 3D simulated model of the scene consisting of multiple objects together with a robot operating on them. This allows for rapid synthesis of appearance, depth, and occlusion information from each camera viewpoint. This information is used both for updating the pose estimates and for extracting the low-level visual cues. The visual cues obtained from each camera are efficiently fused back into the single consistent scene representation using a constrained optimization method. The centralized scene representation, together with the reliability measures it enables, simplify the interaction between pose tracking and pose detection across multiple cameras. We demonstrate the robustness of our approach in a realistic manipulation scenario. We publicly release this work as a part of a general ROS software framework for real-time pose estimation, SimTrack, that can be integrated easily for different robotic applications.

In situations where humanoid robots with constrained posture walk through a narrow space (e.g. manufacturing plants and kitchens), shuffling motions that are stepless and possess wide foot supporting area are effective. One of the difficulties of humanoid's shuffle translations is the load distribution between both feet. If sole loads are not distributed appropriately, the humanoid robot cannot maintain target contact states of each foot, and it will result in slipping both feet or falling down. In this paper, we propose Slide Friction Control (S.F.C.): offline pattern generator and Slide Contact Stabilizer (S.C.S.): online controller. First, Slide Friction Control determines reference foot forces and COM trajectories by adjusting sole loads and considering kinematic friction. The appropriate load distribution of S.F.C. enables humanoid robots to maintain target foot contact states. Second, Slide Contact Stabilizer controls each foot by using damping control to realize reference foot forces determined by S.F.C. S.C.S. enables humanoid robots to slide foot smoothly by suppressing friction vibrations. We also take into consideration the dynamic balance of humanoid robots such as previous waking stabilizers. Finally, we demonstrate that the proposed system enables humanoid robot to slide their feet smoothly using a life-sized humanoid robot, HRP-2.

In this paper, a human robot shared control strategy is developed and tested on a Baxter robot. Using the proposed method, the human operator only needs to consider the motion of the end-effector of the manipulator, while the manipulator will avoid obstacle by itself without sacrificing the end effector motion performance. An improved obstacle avoidance strategy based on the joint space redundancy of the manipulator is designed. A dimension reduction method is presented to solve the over defined problem of avoiding velocity to achieve a more efficient use of the redundancy. By employment of an artificial parallel system of the teleoperate manipulator and the task switching weighting factor, the proposed control method enable the robot restoring back to the commanded pose smoothly when the obstacle is removed. By implementing the dimension reduction method, the trajectory of each joint of the manipulator can be controlled at the same time to achieve the restoring task. Thus, the proposed control method can eliminate the impact of the obstacle on the remaining task. Satisfactory experiment results demonstrate the effectiveness of the proposed methods.

Tactile exploration refers to the use of physical interaction to infer object properties. In this work, we study the feasibility of recovering the shape and pose of a movable object from observing a series of contacts. In particular, we approach the problem of estimating the shape and trajectory of a planar object lying on a frictional surface, and being pushed by a frictional probe. The probe, when in contact with the object, makes observations of the location of contact and the contact normal.

This paper introduces a novel way to improve the settling time of transitions between different walking controllers. This improvement is achieved by commanding a sequence of intermediate transitions to the target controller. As a result, the state of the system enters the domain of attraction of the target controller closer to the fixed point of the Poincaré Map. The method is applicable to any walking robot with one degree of underactuation. The problem is expressed as a Markov Decision Process and then solved with Reinforcement Learning. In order to simplify the stability analysis of underactuated walking the Hybrid Zero Dynamics framework is utilized. Another advantage of using the Hybrid Zero Dynamics is the dimensionality reduction of the state representation in the Markov Decision Process. The experimental results suggest that the proposed methodology performs better than a one-step transition for 84.34% of all the considered transitions for a simulated walking robot matching the parameters of RABBIT [1].

This article depicts an algorithm which matches the output of a Lidar with an initial terrain model to estimate the absolute pose of a robot. Initial models do not perfectly fit the reality and the acquired data set can contain an unknown, and potentially large, proportion of outliers. We present an interval based algorithm that copes with such conditions, by matching the Lidar data with the terrain model in a robust manner. Experimental validations using different terrain model are reported to illustrate the performance of the method.

Industrial robots are important when the degree of automation in industry is increased. To enable the use of robots also when the products change rapidly, the programming must be quick and easy to perform. One way to accomplish this is to use lead-through programming, i.e., the user manually guides the robot. This paper presents a sensorless approach, and thus avoids the need for a typically expensive sensor. The method is based on disabling low-level joint controllers combined with gravity compensation. It is reported how the performance can be improved by compensating for friction. Further, a method for detecting small external torques is described, based on the use of the low-level joint controllers with increased integral gain. The lead-through programming is experimentally evaluated using two different industrial robots.

As there have been many attempts for human-robot collaborations, various solutions to collision detection have been proposed in order to deal with safety issues. However, existing methods for collision detection include the usage of skin sensors or joint torque sensors, and cannot be applied to robots without these sensors such as industrial manipulators. In this study we propose a sensorless collision detection method. The proposed method detects the collision between robots and humans by identifying the external torques applied to the robot. Without using extra sensors, we observed the joint friction model and motor current. We have formulated the friction model for the robot by using the dynamics of the robot and the observer based on the generalized momentum. In addition, formulation of the friction model and the identification scheme did not include any use of extra sensors. The performance of the proposed collision detection method was evaluated using a 7 DOF manipulator. The experimental results show that collision can be reliably detected without any extra sensors for any type of robot arm.

Fish possesses a unique sensory organ called the lateral line. The lateral line provides fish with flow-related information. It is accepted that fish can use the lateral line to sense states of its neighbours in schooling behaviors. In this study, we investigate how a focal robotic fish senses the states of its swimming neighbour by using its onboard artificial lateral line system for the first time. Dye flow visualization is used to characterize the large-scale structures of the wake behind a swimming robotic fish. In the experiment, the Reverse Karman Vortex Street generated by the anterior robotic fish was sensed by the artificial lateral line of the focal robot. The results show that the robot's artificial lateral line can detect the beating frequency of its neighbouring robot and the distance between the robots. It is promising that artificial lateral line sensing could become one of the most popular close interaction methods for a group of underwater robots in the near future.

We present a novel learning algorithm especially designed for challenging, large-scale classification problems in mobile robotics. Our method addresses two important aims: first it reduces the required amount of interaction with a human supervisor, which increases the level of autonomy of the learning process. And second, it has the capability to update its internal representation online with every new observed data sample, which makes it adaptive to new environments. The proposed method is based on a combination of two established methods, namely Online Star Clustering and Label Propagation, but it extends and modifies these in such a way that significant shortcomings such as classification inaccuracy and run time inefficiency can be resolved. In experiments on large benchmark data sets, we show that our approach can quickly learn to classify 3D objects with a significantly reduced amount of required ground truth labels for training.

We propose a novel monocular visual inertial odometry algorithm that combines the advantages of EKF-based approaches with those of direct photometric error minimization methods. The method is based on sparse, very small patches and incorporates the minimization of photometric error directly into the EKF measurement model so that inertial data and vision-based surface measurements are used simultaneously during camera pose estimation. We fuse vision-based and inertial measurements almost at the raw-sensor level, allowing the estimated system state to constrain and guide image-space measurements. Our formulation allows for an efficient implementation that runs in real-time on a standard CPU and has several appealing and unique characteristics such as being robust to fast camera motion, in particular rotation, and not depending on the presence of corner-like features in the scene. We experimentally demonstrate robust and accurate performance compared to ground truth and show that our method works on scenes containing only non-intersecting lines.

Human manipulation activity recognition is an important yet challenging task in robot imitation. In this paper, we introduce, for the first time, a novel method for semantic decomposition and recognition of continuous human manipulation activities by using on-line learned individual manipulation models. Solely based on the spatiotemporal interactions between objects and hands in the scene, the proposed framework can parse not only sequential and concurrent (overlapping) manipulation streams but also basic primitive elements of each detected manipulation. Without requiring any prior object knowledge, the framework can furthermore extract object-like scene entities that are performing the same role in the detected manipulations. The framework was evaluated on our new egocentric activity dataset which contains 120 different samples of 8 single atomic manipulations (e.g. Cutting and Stirring) and 20 long and complex activity demonstrations such as “making a sandwich” and “preparing a breakfast”. We finally show that parsed manipulation actions can be imitated by robots even in various scene contexts with novel objects.