IROS 2016 Papers

We propose a fluidic monostable multivibrator (fMMV) circuit, which is a new flow-switching mechanism in microfluidic chips. Flow control using parallel fMMV circuits does not require an external controller; this is a distinct advantage over various conventional micro-flow control methods. The fMMV circuit mainly consists of an inverter and an RC circuit. Each of the parts uses a microchannel as a resistor, a membrane chamber as a capacitor and a membrane valve as a transistor. The inverter acts as a trigger which opens the membrane valve and the RC circuit delays the closing of the valve, so that flow switching can occur over a time interval. A serial fMMV circuit can switch flows in multiple channels and regulate flowrates with pre-defined durations. By modifying the resistances in fMMV circuits, the flow rates and duration of each phase could be adjusted. For a proof of concept demonstration, we performed experiments using 3 parallel fMMV and 4 parallel fMMV circuit systems and these experiments show the applicability of fMMV circuit to various biochemical processes.

Sampling-based techniques are often employed to solve various complex motion planning problems -the problem of computing a valid path under various robot and/or obstacle constraints. As these methods are random in nature, the probability of their success is directly related to the expansiveness, or openness, of the underlying planning space. However, little is known theoretically in qualifying the conditions under which user (human)-guided approaches improve the efficiency of sampling-based planners. In this paper, we classify and create simplistic models of common user-guided approaches, and we extend the concept of expansiveness to analyze these models to understand both when and how much user-guidance aids sampling-based planners.

In this paper, we try to draw attention to a particular type of force that has been somehow overlooked in the control of mechanical manipulators: gyroscopic force. At first, we broaden our understanding of gyroscopic effect intrinsic in mechanical systems by reformulating the Euler-Lagrange (EL) equation in terms of a gyroscopic part and the non-gyroscopic ones (or Rayleigh type). Then, we propose the use of artificial gyroscopic force to assist existing control laws. As a force that does not do any work, the gyroscopic force can be effectively used in combination with many existing control laws for mechanical systems without affecting their salient features such as the stability or the convergence to a target point. As a specific example, we look into the potential energy shaping parameterized by taskspace variables (also known as the Jacobian transpose method) as a base control law, and combine it with an artificial gyroscopic forcing term. We propose one specific way to design the gyroscopic force as a term quadratic in velocity that operates in reference to a desired velocity field shape (also parameterized in taskspace). The resulting control strategy enables us to shape the intermediate path profiles and thus realizes a taskspace control law that is free from any inverse transformations. We demonstrate the effectiveness of this method using simulation results with a three-link planar manipulator.

This paper studies the effect of a flexible linear torso on the dynamics of passive quadruped bounding. A reduced-order passive and conservative model with translational spinal joint and springy legs is introduced. Numerical return map studies reveal the existence of a confined area of fixed points generating high-speed cyclic bounding motions by exploiting the natural dynamics of the model. The corresponding motion features extensive bidirectional spine deformation. The model displays interesting techniques that result in high-speed locomotion, such as increased flight phase duration and stride length. The obtained results show that the corresponding robot gaits and the associated performance resemble those of its natural counterparts even though the spinal joint lies beyond the bioinspired regime.

Near-Infrared spectroscopy (NIRS) is an emerging non-invasive brain computer interface (BCI) modality that measures changes in haemoglobin concentrations in the cortical tissue. To date most NIRS studies have used standard multiple subject/session dependent classifiers for neural signal decoding. Such approach is preferable to avoid large degree of variabilities in the acquired data that affects classifier generalization. This study presents a classification algorithm that maintains a good performance under the presence of variability in the NIRS data. It is based on ν- support vector machines and its extensions to a multiple kernel learning framework. Empirical evaluations have shown that through the proposed method one can improve the overall BCI decoding accuracy, and its robustness against the variability in neural data.

We present a tracking system based on ultra-wideband (UWB) radio tranceivers mounted on a robot and a target. In comparison to typical UWB localization systems with fixed UWB tranceivers in the environment we only require instrumentation of the target with a single UWB tranceiver. Our system works in GPS-denied environments and does not suffer from long-term drift and limited fields of view. This paper reports the localization algorithm and implementation details. Additionally, we demonstrate a quantitative evaluation of the accuracy (10cm average position error for a square with side-length of 4m) and application scenarios with a quadrotor flying in close proximity to a person and handling occlusion of the target. Finally, we release our implementation as open-source software.

We introduce an optimization-based control approach that enables a team of robots to cooperatively track a target using onboard sensing. In this setting, the robots are required to estimate their own positions as well as concurrently track the target. Our probabilistic method generates controls that minimize the expected uncertainty of the target. Additionally, our method efficiently reasons about occlusions between robots and takes them into account for the control generation. We evaluate our approach in a number of experiments in which we simulate a team of quadrotor robots flying in three-dimensional space to track a moving target on the ground. We compare our method to other state-of-the-art approaches represented by the random sampling technique, lattice planning method, and our previous method. Our experimental results indicate that our method achieves up to 8 times smaller maximum tracking error and up to 2 times smaller average tracking error than the next best approach in the presented scenarios.

We present an object proposal method which utilizes the 3D data obtained from a depth sensor as well as the color information of images. Our object proposal method is designed to improve the performance of the object detection for a mobile robot equipped with a camera and a laser scanner. Compared to traditional object proposal methods using only 2D images, the proposed method provides much less number of candidate windows for object detection. We show less than 100 object proposal windows per image using the proposed method result in high recall tested on the public dataset. Our method presents object proposals in 3D space as well as in 2D image thus it can further be applied to following tasks for mobile robots such as 3D location and pose estimation of the target object after successful object detection. We validate our method using the real-world object detection dataset for outdoor mobile robots captured during the DRC Finals 2015 and the public dataset for comparison with the previous methods.

The development of reliable and robust visual recognition systems is a main challenge towards the deployment of autonomous robotic agents in unconstrained environments. Learning to recognize objects requires image representations that are discriminative to relevant information while being invariant to nuisances, such as scaling, rotations, light and background changes, and so forth. Deep Convolutional Neural Networks can learn such representations from large web-collected image datasets and a natural question is how these systems can be best adapted to the robotics context where little supervision is often available. In this work, we investigate different training strategies for deep architectures on a new dataset collected in a real-world robotic setting. In particular we show how deep networks can be tuned to improve invariance and discriminability properties and perform object identification tasks with minimal supervision.

We present an object detection and tracking framework integrated into a simultaneous localization and mapping (SLAM) system using an RGB-D camera. We propose a compact representation of objects by grouping features hierarchically. Similar to a keyframe being a collection of features, an object is represented as a set of segments, where a segment is a subset of features in a frame. Just like keyframes, segments are registered with each other in a map, which we call an object map. We use the same SLAM procedure in both offline object scanning and online object detection modes. In the offline scanning mode, we scan an object using an RGB-D camera to generate an object map. In the online detection mode, a set of object maps for different objects is given, and the objects are detected via appearance-based matching between the segments in the current frame and in the object maps. In the case of a match, the object is localized with respect to the map being reconstructed by the SLAM system by a RANSAC registration. In the subsequent frames, the tracking is done by predicting the poses of the objects. We also incorporate constraints obtained from the objects into bundle adjustment to improve the object pose estimation accuracy as well as the SLAM reconstruction accuracy. We demonstrate our technique in an object picking scenario using a robot arm. Experimental results show that the system is able to detect and pick up objects successfully from different viewpoints and distances.

Automatic welding of tubular TKY joints is an important and challenging task for the marine and offshore industry. In this paper, a framework for tubular joint detection and motion planning is proposed. The pose of the real tubular joint is detected using RGB-D sensors, which is used to obtain a real-to-virtual mapping for positioning the workpiece in a virtual environment. For motion planning, a Bi-directional Transition-based Rapidly exploring Random Tree (BiTRRT) algorithm is used to generate trajectories for reaching the desired goals. The complete framework is verified with experiments, and the results show that the robot welding torch is able to transit without collision to desired goals which are close to the tubular joint.

Differential flatness is a property of certain systems that greatly simplifies the generation of optimal and dynamically feasible trajectories. Using a differentially flat model, there is no need to integrate the system dynamics to retrieve the states and the constraints of the optimization problem are simpler. Recently, the concept of partial differential flatness has been introduced covering a broader class of systems. In particular, it allows to reduce the need for integration by limiting it to a subset of the states. However, finding an analytical expression for the (partial) differential flatness requires the manipulation of the equations of motion in a very specific manner such that a series of properties are fulfilled. In general, finding such analytical model is not straightforward nor compatible with algorithmic models. In order to tackle this problem, in this paper we present a numerical method to find a (partially) differentially flat model of a system around a collection of states and inputs trajectories. We present results on three underactuated nonlinear systems (cart-pole, planar ballbot and a 3D quadrotor). As use case examples, we show online trajectory re-planning tasks. The validity of the trajectories obtained with the locally flat models is verified by forward integrating the original equations of motion together with an optimal stabilizer.

All the people have their own individual force ranges to feel the sense of touch and its minimum value is called light touch threshold (in short LTT). For instance, if a touching force is very weak, people do not know whether the object is contacted to their skin, even though it is pushing the skin already. This paper presents a novel apparatus to measure the LTT felt by each subject, which is referred to as Active von Frey (AvF) in this paper. As far as the authors know, the LTT measurement device is for the first time proposed in this paper. It is possible for the AvF to provide the touching force ranges from 1[mgf] to 400[mgf]. In order to provide an accurate touching force for the subject, D'Arsonval movement is chosen as an actuator to rotate a touching AvF pin. Both an electric current applied to the D'Arsonval movement and a rotational angle of AvF pin are utilized to calculate the touching force by the electro-mechanical statics. In addition, the image processing technique is used to measure the rotational angle of AvF pin. Finally we show that the LTTs are very different from individual to individual through the experimental results of 10 participants. We hope that the personalized parameters is used to design or control haptic or tele-operation devices for various applications.

Sensor models, which specify the distribution of sensor observations, are a widely used and integral part of robotics algorithms. Observation distributions are commonly approximated by parametric models, which are limited in their expressiveness, and may require careful design to suit an application. In this paper, we propose nonparametric distribution regression as a procedure to model sensors. It is a data-driven procedure to predict distributions that makes few assumptions. We apply the procedure to model raw distributions from real sensors, and also demonstrate its utility to a mobile robot state estimation task. We show that nonparametric distribution regression adapts to characteristics in the training data, leading to realistic predictions. The same procedure competes favorably with baseline parametric models across applications. The results also help develop intuition for different sensor modeling situations. Our procedure is useful when distributions are inherently noisy, and sufficient data is available.

Human locomotion and activity recognition systems form a critical part in a robot's ability to safely and effectively operate in a environment populated with human end users. Previous work in this area relies upon strong assumptions about the labels in the training data; e.g. that are noise-free and that they exist at all. Our approach does not predefine the relevant behaviours or their number, as both are learned directly from observations, similar to real-world human-robot interactions, where labels are neither available. Instead we introduce models that make no assumptions about the state space, by presenting a fully unsupervised nonparametric Bayesian recognition approach, in which we leverage recent advances in state space modelling with automatic inference using probabilistic programming. We demonstrate the utility of full model optimisation using Bayesian optimisation and validate our approach on several challenging problems, using different feature modalities.

This paper presents a practical control approach for series elastic actuators(SEAs) to generate the desired torque. Specifically, the controller is applicable to both linear and nonlinear SEAs and it works well even in the presence of unknown payload parameters and external disturbances. Via the analysis and transformation of the SEA dynamics, a lumped disturbance signal is constructed and a form convenient for controller design is developed. Then, a nonlinear disturbance observer(NDOB) and a sliding-mode control scheme are introduced to synthesize the control law. The performance of the proposed controller is theoretically ensured by Lyapunov analysis. Taking a nonlinear SEA for instance, a series of simulations and hardware experiments are carried out. The results suggest the effectiveness and superior performance of the proposed method for SEA torque control by comparing it with the cascade-PID controller.

In this paper, we demonstrate velocity-level closed-loop control of a tethered magnetic capsule endoscope that is actuated via serial manipulator with a permanent magnet at its end-effector. Closed-loop control (2 degrees-of-freedom in position, and 2 in orientation) is made possible with the use of a real-time magnetic localization algorithm that utilizes the actuating magnetic field and thus does not require additional hardware. Velocity control is implemented to create smooth motion that is clinically necessary for colorectal cancer diagnostics. Our control algorithm generates a spline that passes through a set of input points that roughly defines the shape of the desired trajectory. The velocity controller acts in the tangential direction to the path, while a secondary position controller enforces a nonholonomic constraint on capsule motion. A soft nonholonomic constraint is naturally imposed by the lumen while we enforce a strict constraint for both more accurate estimation of tether disturbance and hypothesized intuitiveness for a clinician's teleoperation. An integrating disturbance force estimation control term is introduced to predict the disturbance of the tether. This paper presents the theoretical formulations and experimental validation of our methodology. Results show the system's ability to achieve a repeatable velocity step response with low steady-state error as well as ability of the tethered capsule to maneuver around a bend.

Stochastic search algorithms are black-box optimizer of an objective function. They have recently gained a lot of attention in operations research, machine learning and policy search of robot motor skills due to their ease of use and their generality. Yet, many stochastic search algorithms require relearning if the task or objective function changes slightly to adapt the solution to the new situation or the new context. In this paper, we consider the contextual stochastic search setup. Here, we want to find multiple good parameter vectors for multiple related tasks, where each task is described by a continuous context vector. Hence, the objective function might change slightly for each parameter vector evaluation of a task or context. Contextual algorithms have been investigated in the field of policy search, however, the search distribution typically uses a parametric model that is linear in the some hand-defined context features. Finding good context features is a challenging task, and hence, non-parametric methods are often preferred over their parametric counter-parts. In this paper, we propose a non-parametric contextual stochastic search algorithm that can learn a non-parametric search distribution for multiple tasks simultaneously. In difference to existing methods, our method can also learn a context dependent covariance matrix that guides the exploration of the search process. We illustrate its performance on several non-linear contextual tasks.

In this paper, we present a fast, non-iterative approach to smooth a noisy input on the Special Euclidean Group, SE(3) manifold. The translational part can be smoothed by a simple Gaussian convolution. We then proposed a novel approach to rotation smoothing. Unlike existing rotation smoothing methods using either iterative optimization methods or stochastic filtering methods, our method allows direct computation of the smoothing result and allows parallelization of the computation. Furthermore, we have done a comparative study on Jia and Evans's method published in 2014 [1], and shown that our method can better smooth an input rotation sequence, with shorter computational time. The smoothed camera path is then used for video stabilisation, which shows fluid and smooth camera motion.