IROS 2016 Papers

In recent years, several researchers focused their work on the mechanisms to protect the UAV from the dangerous collision with obstacles particularly the application of a spherical shell. However, this mechanism has some drawbacks when used in real-world mission especially in an outdoor environment. In the presence of wind, the spherical shell will experience significant air drag that will affect the flight performance of the UAV.

Recent years have seen a growing interest in the use of deep neural networks as function approximators in reinforcement learning. In this paper, an experience replay method is proposed that ensures that the distribution of the experiences used for training is between that of the policy and a uniform distribution. Through experiments on a magnetic manipulation task it is shown that the method reduces the need for sustained exhaustive exploration during learning. This makes it attractive in scenarios where sustained exploration is in-feasible or undesirable, such as for physical systems like robots and for life long learning. The method is also shown to improve the generalization performance of the trained policy, which can make it attractive for transfer learning. Finally, for small experience databases the method performs favorably when compared to the recently proposed alternative of using the temporal difference error to determine the experience sample distribution, which makes it an attractive option for robots with limited memory capacity.

Safe and effective telemanipulation of concentric tube robots is hindered by their complex, non-intuitive kinematics. Guidance schemes in the form of attractive and repulsive constraints can simplify task execution and facilitate natural operation of the robot by clinicians. The real-time seamless calculation and application of guidance, however, requires computationally efficient algorithms that solve the non-linear inverse kinematics of the robot and guarantee that the commanded robot configuration is stable and sufficiently away from the anatomy. This paper presents a multi-processor framework that allows on-the-fly calculation of optimal safe paths based on rapid workspace and roadmap pre-computation The real-time nature of the developed software enables complex guidance constraints to be implemented with minimal computational overhead. A user study on a simulated challenging clinical problem demonstrated that the incorporated guiding constraints are highly beneficial for fast and accurate navigation with concentric tube robots.

The property of the Smoothed Particle Hydrodynamics (SPH) method of being mesh free, adaptable and sui table for tracking of individual particles makes it appropriate for approximating swarm behaviors for multi-agent robotics applications. The researchers modeled each of the swarm robots as SPH particles and then subjected them to external forces to exhibit aggregation and force certain formations. The external forces subjected to the SPH particles are gravity forces and container constraints . The containers explored in the study are simple geometrical primitives: sphere and cube . Computer simulations were done to show how SPH can facilitate in forcing swarm formations with the help of bounding primitives. Algorithm benchmarking was done to show how well SPH performs. Results show that SPH performs better than the benchmark algorithm by a margin of 0.703 and 1.016 units for the two set-ups, respectively. Actual robot implementation was also done to verify the effectivity and viability of the proposed algorithm in exhibiting the aggregation behavior. After 15 seconds of system run time, the interparticle distance and motion accuracy reached 96.93% and 91.14%, respectively.

This paper describes the implementation of a twisting skill to a multi-DOF end effector for manipulating linear deformable objects (LDOs). When manipulating of LDOs such as rope knotting, an important factor is to control torsion. However, controlling torsion is difficult using only the robot arm movement because of the range limit of a manipulator and the danger of collision with the environment such as a table and objects to be tied. On the other hand, implementing the movement into a robot hand as in-hand manipulation means that the robot-arm movement can be compact. In this study, we extract what is needed in a “twisting skill”, implement them into multi-DOF end effectors mounted on a dual-arm robot, and evaluate the skill. Then, we demonstrate rope manipulation including a knotting task and verify the effect by using a twisting skill. Our results show that the proposed method can be used to control the twisted angle of the rope with the required precision. Furthermore, using the proposed twisting skill, when placing a rope on a table top, the accuracy and reproducibility are improved compared with the case when the skill is not used. Finally, using the proposed method, clove hitches are successfully tied.

This paper presents an experimental evaluation of physical human-human interaction in lightweight condition using a one degree of freedom robotized setup. It explores possible origins of Physical Human-Human communication, more precisely, the hypothesis of a time based communication. To explore if the communication is correlated to time a, statistical state machine model based on physical Human-Human interaction is proposed. The model is tested with 14 subjects and presents results that are close to human-human performances.

Series elastic actuators (SEAs) are growingly important in physical human-robot interaction (HRI) due to their inherent safety and compliance. Cable-driven SEAs also allow flexible installation and remote torque transmission, etc. However, there are still challenges for the impedance control of cable-driven SEAs, such as the reduced bandwidth caused by the elastic component, and the performance balance between reference tracking and robustness. In this paper, a velocity sourced cable-driven SEA has been set up. Then, a stabilizing 2 degrees of freedom (2-DOF) control approach was designed to separately pursue the goals of robustness and torque tracking. Further, the impedance control structure for human-robot interaction was designed and implemented with a torque compensator. Both simulation and practical experiments have validated the efficacy of the 2-DOF method for the control of cable-driven SEAs.

In this paper, an impedance control scheme for aerial robotic manipulators is proposed, with the aim of reducing the end-effector interaction forces with the environment. The proposed control has a multi-level architecture, in detail the outer loop is composed by a trajectory generator and an impedance filter that modifies the trajectory to achieve a complaint behaviour in the end-effector space; a middle loop is used to generate the joint space variables through an inverse kinematic algorithm; finally the inner loop is aimed at ensuring the motion tracking. The proposed control architecture has been experimentally tested.

This work addresses path planning for reactionless visual servoing of a redundant dual-arm space robot through exploration in the image space. The planner explores the image moment based feature space, impends acceleration to the image features and extends the feature tree. A reactionless visual servoing control law is integrated to extend the tree in the configuration space simultaneously. The proposed algorithm is able to incorporate the necessary coupling between the motions of the the dual arms and the base of the robot to ensure zero base reactions. Additionally, it also gives the flexibility to apply multiple constraints in both the image space and the configuration space. The effectiveness of the proposed framework is exhibited by implementing the algorithm on a numerical model of a 14-DoF dual arm space robot.

Illumination changes are a typical problem for many outdoor long-term applications such as visual place recognition. Keypoints may fail to match between images taken at the same location but different times of the day. Although recently some methods are presented for creating shadow-free image representations, all of them have the limitation in terms of dealing with night images and non-Planckian source of lighting. In this paper we present a new method for creating illumination invariant image representation using a combination of two existing methods based on natural image statistics that address the issue of illumination invariance. Unlike previous attempts at solving the problem of illumination invariant representation, the proposed method does not assume the ideal narrow-band color camera nor a calibration step for each environment. We evaluate our method on real datasets to establish its accuracy and efficiency. Experimental results show that our method outperforms competing methods for illumination invariant image representation.

This paper presents a new condition, the fully physical consistency for a set of inertial parameters to determine if they can be generated by a physical rigid body. The proposed condition ensure both the positive definiteness and the triangular inequality of 3D inertia matrices as opposed to existing techniques in which the triangular inequality constraint is ignored. This paper presents also a new parametrization that naturally ensures that the inertial parameters are fully physical consistency. The proposed parametrization is exploited to reformulate the inertial identification problem as a manifold optimization problem, that ensures that the identified parameters can always be generated by a physical body. The proposed optimization problem has been validated with a set of experiments on the iCub humanoid robot.

This work deals with interactive motion planning processes intended to assist a human operator when simulating industrial tasks such as assembly, maintenance or disassembly in Virtual Reality. Such applications need motion planning on surfaces. We propose an original interactive path planning algorithm with contact, I-RRT-C, based on a RRT-Connect approach. This algorithm is based on a real-time interactive approach allowing both an automatic motion planner and a human operator to jointly explore the workspace. A parameter balances the authority between the computer and the operator to reduce processing times. We improve the guidance by allowing to sample on the surfaces of obstacles. Our method allows to find a path in cluttered environments or to solve contact operations such as insertion or sliding tasks. Last, we present experimental results showing that our interactive path planner with contact brings a significant improvement over state of the art methods in both free and contact space.

Most legged animals exploit flexible body and supporting muscles to produce power for dynamic behaviors which results in fast locomotion and additional mobility. Previous works have focused on the symmetric flexible body with massless legs associated to the body. However, body bending in animals during running happens prior to the rear side instead of the middle point of body. Therefore, a quadruped model with a passive spinal joint, asymmetric segmented body, actuated hip joints and legs is introduced. By using a numerical return map, a periodic bounding locomotion of the model is found with optimal sets of initial conditions and proper system parameters. Moreover, this paper investigates the effects of spine flexibility in segmented body on quadrupedal bounding gait. The results show that asymmetric segmented body has bigger spine oscillation, shorter stride period and smaller cost of transport, which helps the robot run more efficiently.

During human-robot collaboration tasks, we may physically touch the robot at a generic location and engage an intentional exchange of forces while realizing coordinated motion of the common contact point. In order to control the relative motion and the exchanged contact forces, the latter need to be estimated without using any local force sensing device. Building upon our recent works, we generalize the classical hybrid force/velocity control design to this situation, handling complementary quantities along the directions of a suitable contact task frame in a dynamically decoupled way. The contact force is estimated online using our residual method together with an external sensor to localize the contact point, and the time-varying contact task frame is obtained analytically from this estimate. Experimental results are presented for a KUKA LWR4 robot using a Kinect sensor.

This paper explores combining task and manipulation planning for humanoid robots. Existing methods tend to either take prohibitively long to compute for humanoids or artificially limit the physical capabilities of the humanoid platform by restricting the robot's actions to predetermined trajectories. We present a hybrid planning system which is able to scale well for complex tasks without relying on predetermined robot actions. Our system utilizes the hybrid backward-forward planning algorithm for high-level task planning combined with humanoid primitives for standing and walking motion planning. These primitives are designed to be efficiently computable during planning, despite the large amount of complexity present in humanoid robots, while still informing the task planner of the geometric constraints present in the problem. Our experiments apply our method to simulated pick-and-place problems with additional gate constraints impacting navigation using the DRC-HUBO1 robot. Our system is able to solve puzzle-like problems on a humanoid within a matter of minutes.

Our human-robot collaboration research aims to improve the fluency and efficiency of interactions between humans and robots when executing a set of tasks in a shared workspace. During human-robot collaboration, a robot and a user must often complete a disjoint set of tasks that use an overlapping set of objects, without using the same object simultaneously. A key challenge is deciding what task the robot should perform next in order to facilitate fluent and efficient collaboration. Most prior work does so by first predicting the human's intended goal, and then selecting actions given that goal. However, it is often difficult, and sometimes impossible, to infer the human's exact goal in real time, and this serial predict-then-act method is not adaptive to changes in human goals. In this paper, we present a system for inferring a probability distribution over human goals, and producing assistance actions given that distribution in real time. The aim is to minimize the disruption caused by the nature of human-robot shared workspace. We extend recent work utilizing Partially Observable Markov Decision Processes (POMDPs) for shared autonomy in order to provide assistance without knowing the exact goal. We evaluate our system in a study with 28 participants, and show that our POMDP model outperforms state of the art predict-then-act models by producing fewer human-robot collisions and less human idling time.

Advances in automation have the potential to reduce the workload required for human planning and execution of missions carried out by robotic systems such as unmanned aerial vehicles (UAVs). However, automation can also result in an increase in system complexity and a corresponding decrease in system transparency, which makes identifying and reasoning about errors in mission plans more difficult. To help explain errors in robotic planning systems, we define a notion of structured probabilistic counterexamples, which provide human-interpretable diagnostic information about requirements violations resulting from complex probabilistic robotic behavior. We propose an approach for generating such counterexamples using mixed integer linear programming and demonstrate the usefulness of our approach via a case study of UAV mission planning demonstrated in the AMASE multi-UAV simulator.

Map building is a fundamental task in many robotic applications. In this paper, we propose a novel approach for 3D mapping of indoor environments which allows a robot avatar to collaborate with a human seamlessly through a virtual reality (VR) device. The 3D map is created using the 3D data from an RGB-D camera mounted on the robot and simultaneously transmitted to a remote server, and then rendered to the VR device. On the other hand, the intentions of the user are inferred using the motion of the head movement based on hidden Markov models (HMMs), and then interpreted into commands to control the robot. We implement the proposed approach based on a modified Pioneer robot platform. The experimental results show the feasibility of the proposed system.

We propose a human mimetic foot structure for musculoskeletal humanoids. We designed the foot structure by inspiring from human foot abilities of the multi-bone connected structure for flexibility and the distributed force sensor system. The foot has multi-DOFs structure including toe DOF that is composed of fingers. The distributed force sensing system is composed of 12 an-axis force sensors. In order to demonstrate those effectiveness, we implement the foot into musculoskeletal humanoid Kengoro and conduct several experiments. As a result, we confirmed effectiveness of the foot from tiptoe motion and balancing behavior by utilizing the foot characteristics.