IROS 2015 论文列表

sEMG recognition has been used extensively in prosthetic device control, human-assisting manipulators and sign language recognition, etc. However, the sEMG recognition model, trained with one subject's sEMG data, is not applicable to the other subjects, which hinders the practical application of myoelectric interfaces immensely. In this paper, a sEMG recognition method which is applicable to multi-users is proposed. Firstly, single channel sEMG is decomposed into 30 MUAPTs, which includes four steps: two-order differential filter, threshold calculation, spike detection and hierarchical clustering. Secondly, the MUAPTs are updated with the templates orthogonalization; and Deep Boltzman Machine is employed to classify the MUAPTs into five classes corresponding to the predefined five gestures. Six participants participated in this experiment to validate the effectiveness of the proposed method. Results indicated that this method can achieve a mean accuracy of 81.5%.

Methods from machine learning have successfully been used to improve the performance of control systems in cases when accurate models of the system or the environment are not available. These methods require the use of data generated from physical trials. Transfer Learning (TL) allows for this data to come from a different, similar system. This paper studies a simplified TL scenario with the goal of understanding in which cases a simple, alignment-based transfer of data is possible and beneficial. Two linear, time-invariant (LTI), single-input, single-output systems are tasked to follow the same reference signal. A scalar, LTI transformation is applied to the output from a source system to align with the output from a target system. An upper bound on the 2-norm of the transformation error is derived for a large set of reference signals and is minimized with respect to the transformation scalar. Analysis shows that the minimized error bound is reduced for systems with poles that lie close to each other (that is, for systems with similar response times). This criterion is relaxed for systems with poles that have a larger negative real part (that is, for stable systems with fast response), meaning that poles can be further apart for the same minimized error bound. Additionally, numerical results show that using the reference signal as input to the transformation reduces the minimized bound further.

Series elastic actuators are frequently modeled using a conventional lumped mass model which has remained mostly unchanged since their introduction almost two decades ago. The lumped model has served well for early development but more descriptive models are now needed for new actuator designs and control approaches. In this paper we propose a new unlumped model specifically for linear series elastic actuators which uses a rack & pinion conceptualization to intuitively depict the mechanics of a linear ball screw drive. Results from hardware experiments are presented and compared to the predicted simulation results for both the conventional model and the new unlumped model. The results demonstrate that the new unlumped model is significantly more representative of the true actuator dynamics.

The kinematic modelling has been applied to many controllers of under-actuated manipulators. Most of these studies assume that the control process is conducted within the workspace. However, as such a kinematic model cannot describe the situations when the stable grasping is violated in the real environment, these controllers may fail unexpectedly. In this paper, we propose a combination of kinematics based Workspace Analysis (WA) and Gaussian Process Classification (GPC) to model the success rates of control actions in the theoretical workspace. We also use the Gaussian Process Regression (GPR) to model the residual between the prediction of the WA and the ground truth data. We then apply this integrated model, Gaussian Processes enhanced Workspace Analysis (GP-WA), into an optimal controller. The optimal controller is implemented on a planar under-actuated gripper with two three-phalanx fingers. Two sets of simulation experiments are carried out to validate our method. The results demonstrate that the optimal manipulation controller based on GP-WA achieves high control accuracy for manipulating a wide range of objects.

Magnetic drug targeting is an approach to guide and concentrate magnetic nanoparticles (MNPs) into the diseased target organ after being injected into blood vessels. Although many works for drug targeting have been conducted, there are few studies on delivering the nanoparticles to the target region. Drug delivery performance has not been addressed sufficiently or fully. In this paper, we investigate the effect of dominant factors to MNPs delivery performance. Then, an optimized field function scheme with a pulsed magnetic actuation is proposed to significantly improve the MNPs guidance performance. With a specific condition of blood vessel size, particle size, and applied magnetic field, the optimized parameters of the field function are selected through extensive simulation studies. We find out that the optimal negative and positive time for the magnetic pulsed field mainly depends on the exit time for particles to reach the bifurcation and the critical time as the maximum time for them to reach the vessels wall, respectively. With the chosen parameters, we show that ratios of correctly guided particles in a Y-channel are reached to 100%. In addition, to minimize the power consumption, a modified field function (MFF) scheme is introduced. The MFF includes a no-power time, called zero-time, between the positive and negative time. It is shown that with the proposed MFF, the energy consumption and the heating problem of the actuator system can be significantly reduced. Therefore, the proposed guidance scheme for MNPs can overcome the sticking issue and maximize the guidance performance as well as reducing the power consumption. It should be noted that the MFF can be easily implement by programmable DC power supplies connected to electromagnetic coils.

This paper proposes and implements an auditory perception platform for a home service robot that serves the elderly living alone at home. A layered architecture for the proposed platform was developed to realize the various auditory perception capabilities while enabling a remote caregiver to involve in the sound event recognition process. We successfully implemented the software for this architecture that realizes robot services and auditory services for developing high level auditory applications. The robot is able to estimate the sound source position and recognize human speech in the room with multiple sound sources, as well as to collaborate with the caregiver on sound event recognition. Our experimental results validated the proposed platform.

In this paper, we introduce an approach for leveraging CAD description to a semantic level, in order to link additional knowledge to CAD models and to exploit resulting synergy effects.

When a robotic manipulator is controlled, computing gravity force of the robot is the primary issue. Exact model parameters are not easy to be known in the practical robot system due to the uncertainty of the robot dynamics. Hence, the gravity force is presented by a combination of gravity regressor and robot dynamic parameters and is compensated by the estimation of uncertain robot dynamic parameters. Previous researches conducted estimation by using transpose of gravity regressor and full form of dynamic parameters which is not general form however this paper estimates the gravity force using the generalized gravity regressor which is regardless of the dimension and structure of the robot under the quasi-static state. Once the estimation is completed, the estimated value can be used to compute the gravitational force and control the robot. It is shown that the generalized decomposition of gravity regressor and estimation process. The results are validated through an experiment by implementing the algorithm on an upper-body dual arm robot.

A novel method for sensing local pressure inside a microfluidic device is proposed and developed. The main advantage of the method is that the pressure can be visually seen without attaching any electrical wire or instrument. The method can also be easily integrated into applications, such as micro-robots, because of its single-layer design. The working principle is based on the deformation of Polydimethylsiloxane, a kind of polymer that is often used in microfluidic systems. The deformation in the deformation chamber causes fluid to flow in, or out, the sensing area, and consequently results in changes of color intensity in the area. As a result, the pressure can be determined based on the brightness value over the sensing area. Experimental results show a highly correlated relationship between the measured brightness and pressure with the absolute correlation of 0.986. The hysteresis and drift of the proposed sensor are investigated, and are compared with our previously proposed microbeads approach. The prototype of the proposed sensor succeeded in performing measurement on local pressure in a microfluidic device with the resolution of 3.04 kPa.

Fiber reinforced elastomeric actuators are popular actuators for soft robots because of their inherent safety, energy density and a large repertoire of spatial motion patterns. However, a small subset of these actuators alone known as McKibben pneumatic muscles with antisymmetric fiber orientations have been extensively analyzed in literature. This paper analyzes the large deformation kinematics of generalized McKibben actuators with asymmetric and arbitrarily varying fiber orientations by formulating a simple and accurate isoperimetric problem that involves constrained volume maximization problem. This model maximally decouples kinematics and kinetostatics thereby significantly reducing the numerical complexity involved in analysis. The accuracy of the model is verified by benchmarking with existing models for the McKibben actuator case, and with experiments for novel designs with no associated prior literature. This model is deemed to be useful in the design synthesis of fiber reinforced elastomeric actuators for a desired kinematic and kinetostatic requirement.

In this work a modification of a Cartesian impedance controller is presented. The proposed controller guarantees compliance of the operated manipulator strictly within preselected spatial limits. Specifically, by introducing a nonlinear stiffness term into the control design, the manipulator is prevented from colliding with an explicitly defined hard boundary, like a human being or a fragile object. When operating in spatial regions sufficiently away from the boundary the manipulator behaves as if it were operated via a linear stiffness impedance controller, thus preserving in these regions its compliance and convergence characteristics. A comparative simulation with a linear stiffness impedance controller is performed to verify and clarify the proposed approach.

The authors have investigated a construction methodology for a non-invasive ultrasound theragnostic system, called NIUTS, to track and follow moving body targets, such as tumors / stones located in moving organs (kidneys, livers, etc.), robustly and precisely, so as to compensate for nearly 90 percent of target motions in ultrasound images, while irradiating therapeutic ultrasound to focal lesions. In the present report, we propose a novel servo recovery algorithm for NIUTS. We confirm the effectiveness of the implemented servo recovery algorithm for a real human kidney, which, as far as we know, is the first successful report of semi-autonomous servo recovery systems for lost body targets. This robust servo recovery method could provide an indispensable tool for precise, safe, and efficient treatments of tumors and stones.

This paper presents a control method for walking assist with hip-mounted exoskeleton robots. For modeling a user's current walking motion, a novel finite state machine is first constructed. We divide a walking cycle uniformly using the inevitable zero crossing events. When state transitions occur, we capture the current walking spatio-temporal sensor data as discrete form. By using the sensed hip data as boundary conditions, we also develop a gait primitives based motion reconstruction method. Gait primitives are a form of basis trajectories to represent various joint motions. From those methods we estimate the moment of heel landing with interpolated knee joint motions. Utilizing the user's previous opposite step motion, we predict the positive or negative work intervals of the current step motion. This makes it possible to achieve natural ‘one shot’ assist by driving adapted torques fast. This assist strategy is also effective to enhance gait regularity. The measures of stride time variability are improved by over 30% for the simulated experiment. Various real experimentations demonstrate the feasibility of our approach.

Keyframe-based Learning from Demonstration has been shown to be an effective method for allowing end-users to teach robots skills. We propose a method for using multiple keyframe demonstrations to learn skills as sequences of positional constraints (c-keyframes) which can be planned between for skill execution. We also introduce an interactive GUI which can be used for displaying the learned c-keyframes to the teacher, for altering aspects of the skill after it has been taught, or for specifying a skill directly without providing kinesthetic demonstrations. We compare 3 methods of teaching c-keyframe skills: kinesthetic teaching, GUI teaching, and kinesthetic teaching followed by GUI editing of the learned skill (K-GUI teaching). Based on user evaluation, the K-GUI method of teaching is found to be the most preferred, and the GUI to be the least preferred. Kinesthetic teaching is also shown to result in more robust constraints than GUI teaching, and several use cases of K-GUI teaching are discussed to show how the GUI can be used to improve the results of kinesthetic teaching.

We present a computationally efficient RGB-D based pose estimation solution for less computationally resourced MAVs, which are ideally suited as members in a swarm. Our approach applies the sufficient statistics derived for a least-squares problem to our problem context. RANSAC-based outlier detection in aligning corresponding feature points is a time consuming operation in visual pose estimation. The additive nature of the used sufficient statistics significantly reduces the computation time of the RANSAC procedure since the pose estimation in each test loop can be computed by reusing previously computed sufficient statistics. This eliminates the need for recomputing estimates from scratch each time. A simpler hypotheses testing method gave similar performance in terms of speed but less accurate than our proposed method. We further increase the efficiency by reducing the problem size to four dimensions using attitude data from an Attitude and Heading Reference System (AHRS). Using a real-world dataset, we show that our algorithm saves up to 94% of computation time for the RANSAC-based procedure in pose estimation while improving the accuracy.

Bi-directional search is a widely used strategy to increase the success and convergence rates of sampling-based motion planning algorithms. Yet, few results are available that merge both bi-directional search and asymptotic optimality into existing optimal planners, such as PRM*, RRT*, and FMT*. The objective of this paper is to fill this gap. Specifically, this paper presents a bi-directional, sampling-based, asymptotically-optimal algorithm named Bi-directional FMT* (BFMT*) that extends the Fast Marching Tree (FMT*) algorithm to bi-directional search while preserving its key properties, chiefly lazy search and asymptotic optimality through convergence in probability. BFMT* performs a two-source, lazy dynamic programming recursion over a set of randomly-drawn samples, correspondingly generating two search trees: one in cost-to-come space from the initial configuration and another in cost-to-go space from the goal configuration. Numerical experiments illustrate the advantages of BFMT* over its unidirectional counterpart, as well as a number of other state-of-the-art planners.

This paper presents a distributed world model that is able to adapt to changes in the Quality of Service (QoS) of the communication layer by online reconfiguration of perception algorithms. The approach consists of (a) a mechanism for storage, exchange and processing of world model data and (b) a feedback loop that incorporates reasoning techniques to adapt to QoS changes immediately. The latter introduces a Level of Detail (LoD) metric based on a spatial resolution in order to infer an upper bound for the amount of data that can be transmitted without violating an application specific transmission delay. Experiments have been performed with Octree-based subsampling techniques applied to data originating from a RGB-D camera using simulated and real-world data sets for timevarying bandwidth values as employed QoS measure.

This paper introduces an adaptive tuning method for the controllers of a 4 degrees-of-freedom autonomous underwater vehicle. The proposed scheme consists of two control loops, one for position control and an inner one for velocity control. The gains of the controller are determined on-line, according to the position/velocity errors. Using the proposed adaptive architecture, the uncertainties in the parameters of the system are addressed and the system is able to operate when hydrodynamic disturbances are present. The complexity of the fixed gain tuning procedure is also greatly decreased for underwater vehicles when the algorithm suggested here is used. Experimental results with the Nessie VII AUV show that the adaptive controller is beneficial for underwater vehicles. Finally it is shown that the current approach reduces the energy consumption of the system.

This paper illustrates a method to design dynamic neural network controllers to allow a robot to autonomously navigate roads based on color perception. The neuro-controller moves the robot by setting the speed of the wheels and adjusts the robot visual systems by setting the value of three parameters that determine how much of the red, green and blue components of the RGB camera images contribute in generating the network input vector. Results show that the best evolved controller can successfully drive a real robot in environments with color characteristics never encountered during evolution. Moreover, we show that the dynamic color perception abilities are based on complex patterns of activation of the three color parameters. These patterns are generated by evolved neural mechanisms that successfully adapt the robot perceptual system to the color characteristics of the different visual scenes.