ICRA 2015 论文列表

The tiny nematode worm Caenorhabditis elegans has long been a popular model organism for genetic, developmental, and biochemical studies in which worm microinjection plays a critical role. This paper presents an automated robotic system for high-speed injection of C. elegans with an efficiency more than 10 times faster than that of a proficient injection technician. To facilitate the injection process, a multilayer, hydraulically-controlled polydimethylsiloxane (PDMS) microfluidic device is developed to rapidly load, immobilize, flush, sort and collect individual worms. In addition, a newly proposed contact detection algorithm is adopted to find the optimal injection position along the z axis within the microscope view field. The direction and location of the needle tip are identified online based on an effective image processing algorithm. According to continuous injection of 40 worms, our system is able to perform worm injection at a speed of 6.6 worms per minute with a pre-sorting success rate of 77.5% (post sorting: 100%). The superior performance provided by the system will significantly facilitate large-scale transgenic studies and biomolecule screening on C. elegans.

Supervisory control techniques (SCTs), as one of the fundamental researches in the domain of automated manufacturing systems, implement specifications upon a plant model, leading the system to tendentiously desired behaviors. Generalized Mutual Exclusion Constraints (GMECs) are a typical kind of SCTs which correspond to a linear supervisory specification in the context of Petri nets (PNs). Given a set of admissible GMECs, supervisor simplification is needed to reduce implementation cost and mitigate fabrication complexity. In this paper, two methods are proposed and improved in the paradigm of arbitrary GMECs simplification. For the method based on P-invariant control, it develops in the semantics of PN structure analysis, exhibiting drawbacks like low computational efficiency, high failure probability, and definite failure to control weakly L1-dependent inequalities. On the contrary, for the method based on inequality analysis, it germinates on the ground of algebra theory, tackling both L1- and L2-dependent inequalities. Experimental results show the effectiveness and efficiency of supervisor simplification based on inequality analysis.

There are many benefits for the deployment of multiple autonomous industrial robots to carry out a task, particularly if the robots act in a highly collaborative manner. Collaboration can be possible when each robot is able to autonomously explore the environment, localize itself, create a map of the environment and communicate with other robots. This paper presents an approach to the modeling of the collaboration problem of multiple robots determining optimal base positions and orientations in an environment by considering the team objectives and the information shared amongst the robots. It is assumed that the robots can communicate so as to share information on the environment, their operation status and their capabilities. The approach has been applied to a team of robots that are required to perform complete surface coverage tasks such as grit-blasting and spray painting in unstructured environments. Case studies of such applications are presented to demonstrate the effectiveness of the approach.

Ankle inversion-eversion compliance is an important feature of conventional prosthetic feet, and control of inversion, or roll, in robotic prostheses could improve balance for people with amputation. We designed a tethered ankle-foot prosthesis with two independently-actuated toes that are coordinated to provide plantarflexion and inversion-eversion torques. This configuration allows a simple lightweight structure with a total mass of 0.72 kg. Strain gages on the toes measure torque with less than 2.7% RMS error, while compliance in the Bowden cable tether provides series elasticity. Benchtop tests demonstrated a 90% rise time of less than 33 ms and peak torques of 180 N·m in plantarflexion and ±30 N·m in inversion-eversion. The phase-limited closedloop torque bandwidth is 20 Hz with a 90 N·m amplitude chirp in plantarflexion, and 24 Hz with a 20 N·m amplitude chirp in inversion-eversion. The system has low sensitivity to toe position disturbances at frequencies of up to 18 Hz. Walking trials with five values of constant inversion-eversion torque demonstrated RMS torque tracking errors of less than 3.7% in plantarflexion and less than 5.9% in inversion-eversion. These properties make the platform suitable for haptic rendering of virtual devices in experiments with humans, which may reveal strategies for improving balance or allow controlled comparisons of conventional prosthesis features. A similar morphology may be effective for autonomous devices.

Due to challenges, such as information collection in huge water, propulsion at different scales and manipulation in extreme fluid environments, wide application of unmanned underwater vehicles (UUVs) is limited. In this paper, an alligator modular robot, composed by 14 small robots, was inspired from the locomotion and morphologies of alligators and microorganisms. With the undulatory body and beating legs, the assembled large robot was able to maintain fast and maneuverable swimming for tasks executions at large Reynolds number. With a biologically optimized body aspect ratio (3.47), small modular robots achieved effective propulsion in viscous fluid. Facilitated by a dissembling system, small robots were automatically released from the large robot frame and swam around with independent actuation systems. Hydrodynamic model for the large and small robot were formulated, based on which the propulsion for the modular robot in large and low Reynolds numbers (Res) were simulated. Experiments were conducted to compare with the simulation results to further validate the proposed modular robot design.

This paper presents an adaptive and compliant approach-to-grasp strategy for multi-finger robotic hands, to improve the performance of autonomous grasping when encountering object position uncertainties. With the proposed approach-to-grasp strategy, the first robot finger to experience unexpected impact would pause its movement in a compliant manner, and remains in contact with the object to minimize the unplanned motion of the target object. At the same time, the remainder of the fingers continuously, adaptively move toward re-adjusted grasping positions with respect to the first finger in contact with the object, without the need for on-line re-planning or re-grasping. An adaptive grasp control strategy based on spatial virtual spring framework is proposed to achieve local (e.g. not resorting to the robotic arm) in-hand adjustments of the fingers not yet in contact. As such, these fingers can be adaptively driven to the adjusted desired position to accomplish the grasp. Experimental results demonstrate that significantly larger position errors with respect to the hand workspace can be accommodated with the proposed adaptive compliant grasp control strategy. As much as 391% increase in position error area coverage has been achieved. Finally, beyond the quantitative analysis, additional observations during the extensive experiment trials are discussed qualitatively, to help examine several open issues, and further understand the approach-to-grasp phases of the robot hand tasks.

This paper investigates the use of single control signal (magnetic field direction) and PSO-MPC algorithm to control multiple magnetized Tetrahymena pyriformis (T. pyriformis) cells to move from their initial positions to their target positions simultaneously while avoiding the obstacle. The magnetized T. pyriformis cells are generated by adding iron-oxide spherical particles into the cells. We control the cells' moving direction by changing the magnetic field direction. Based on Model Predictive Control (MPC) algorithm, we define a cost function which is composed of the target cost function and the obstacle potential function. The target cost function is to measure the sum of differences between cells' predicted positions and their target positions. The obstacle potential function is used to measure the repulsive force of the obstacle. The input variables of the cost function are the sequence of control signals. We use Particle Swarm Optimization (PSO) method to find a cost value which is close to the global minimum of the cost function. In the experimental result section, we show the control of three m3pi robots to move from their initial positions to their target positions with avoiding the obstacle. Since the similar control strategy has successfully controlled one T. pyriformis cell in our previous work, we believe our PSO-MPC algorithm is applicable on the multiple T. pyriformis cells' control task.

We present an algorithm that uses a sparse collection of noisy sensors to characterize the observed behavior of a mobile agent. Our approach models the agent's behavior using a collection of randomized simulators called implicit agent models and seeks to classify the agent according to which of these models is believed to be governing its motions. To accomplish this, we introduce an algorithm whose input is an observation sequence generated by the agent, represented as sensor label-time pairs, along with an observation sequence generated by one of our implicit agent models and whose output is a measure of the similarity between the two observation sequences. Using this similarity measure, we propose two algorithms for the model classification problem: one based on a weighted voting scheme and one that uses intermediate resampling steps. We have implemented these algorithms in simulation, and present results demonstrating their effectiveness in correctly classifying mobile agents.

As robots begin to collaborate with humans in everyday workspaces, they will need to understand the functions of tools and their parts. To cut an apple or hammer a nail, robots need to not just know the tool's name, but they must localize the tool's parts and identify their functions. Intuitively, the geometry of a part is closely related to its possible functions, or its affordances. Therefore, we propose two approaches for learning affordances from local shape and geometry primitives: 1) superpixel based hierarchical matching pursuit (S-HMP); and 2) structured random forests (SRF). Moreover, since a part can be used in many ways, we introduce a large RGB-Depth dataset where tool parts are labeled with multiple affordances and their relative rankings. With ranked affordances, we evaluate the proposed methods on 3 cluttered scenes and over 105 kitchen, workshop and garden tools, using ranked correlation and a weighted F-measure score [26]. Experimental results over sequences containing clutter, occlusions, and viewpoint changes show that the approaches return precise predictions that could be used by a robot. S-HMP achieves high accuracy but at a significant computational cost, while SRF provides slightly less accurate predictions but in real-time. Finally, we validate the effectiveness of our approaches on the Cornell Grasping Dataset [25] for detecting graspable regions, and achieve state-of-the-art performance.

Homeland safety and security will increasingly depend upon autonomous unmanned vehicles as a method of assessing and maintaining situational awareness. As autonomous team algorithms evolve toward requiring less human intervention, it may be that having an “operator-in-the-loop” becomes the ultimate goal in utilizing autonomous teams for surveillance. However studies have shown that trust plays a factor in how effectively an operator can work with autonomous teammates. In this work, we study mechanisms that look at autonomy as a system and not as the sum of individual actions. First, we conjecture that if the operator understands how the team autonomy is designed that the user would better trust that the system will contribute to the overall goal. Second, we focus on algorithm input criteria as being linked to operator perception and trust. We focus on adding a time-varying spatial projection of areas in the ROI that have been unseen for more than a set duration (STEC). Studies utilize a custom test bed that allows users to interact with a surveillance team to find a target in the region of interest. Results show that while algorithm training had an adverse effect, projecting salient team/surveillance state had a statistically significant impact on trust and did not negatively affect workload or performance. This result may point at a mechanism for improving trust through visualizing states as used in the autonomous algorithm.

This paper presents an aerial robot with a manipulator to implement door opening mission. Although general aerial robots have advantages of flying in the three-dimensional space freely, they don't have any capability of moving to another room when the door is closed. To overcome this problem, we propose a new configuration of an aerial manipulator with perching function, knob-twisting function, and door-pushing function. Perching function can be achieved by integration of door-approach control, the mechanism for perching and attitude-change control. With regard to knob-twisting function, the design concept of a light-weight manipulator generating large enough force to twist the knob is introduced, which is composed of a soft-bag actuator with variable restriction to perform the arbitrary curved trajectory. On the other hand, the door pushing force is aimed to be generated by the lift of the propeller, which is helpful to avoid gaining the additional weight. The validity of the proposed methods is experimentally verified by using the developed prototype.

Controlling mobile robots with complex articulated parts and hence many degrees of freedom generates high cognitive load on the operator, especially under demanding conditions such as in Urban Search & Rescue missions. We propose a solution based on reinforcement learning in order to accommodate the robot morphology automatically to the terrain and the obstacles it traverses. In this paper, we concentrate on the crucial issue of predicting rewards from incomplete or missing data. For this purpose we exploit the Gaussian processes as a predictor combined with decision trees. We demonstrate our achievements in a series of experiments on real data.

Saccades are fast eye movements that allow humans and robots to bring the visual target in the center of the visual field. Saccades are open loop with respect to the vision system, thus their execution require a precise knowledge of the internal model of the oculomotor system. In this work, we modeled the saccade control, taking inspiration from the recurrent loops between the cerebellum and the brainstem. In this model, the brainstem acts as a fixed-inverse model of the oculomotor system, while the cerebellum acts as an adaptive element that learns the internal model of the oculomotor system. The adaptive filter is implemented using a state-of-the-art neural network, called I-SSGPR. The proposed approach, namely recurrent architecture, was validated through experiments performed both in simulation and on an antropomorphic robotic head. Moreover, we compared the recurrent architecture with another model of the cerebellum, the feedback error learning. Achieved results show that the recurrent architecture outperforms the feedback error learning in terms of accuracy and insensitivity to the choice of the feedback controller.

The prominent maneuverability of flapping flight is enabled by rapid and significant changes in aerodynamic forces, which is a result of surprisingly subtle and precise changes of wing kinematics. The high sensitivity of aerodynamic forces to wing kinematic changes demands precise and instantaneous control of the flapping wing trajectories, especially in the presence of various types of uncertainties. In this work, we first present a dynamic model of a pair of direct-motor-driven flapping wings while taking into consideration the parameter uncertainties and external disturbances. We then present an Adaptive Robust Controller (ARC) to achieve robust performance of high-frequency (over 30Hz) instantaneous wing trajectory tracking with onboard feedback. The proposed control algorithm was experimentally validated on a 7.5 gram flapping-wing MAV which showed excellent tracking of various wing trajectories with different amplitude, bias, frequency, and split-cycles. Experimental results on various model wings demonstrated that the ARC can adapt to unknown parameters and show no performance degradation across wings of different geometries. The results of ARC were also compared with those of open-loop and classical PID controllers.

This paper presents an extension of the standard occupancy grid for 3D environment mapping. The presented approach adds a fusion process after the occupancy update which modifies the resolution of the grid cells in an incremental manner. Consequently, the proposed approach requires fewer grid cells for 3D representation in comparison to a standard occupancy grid. The resolution adaptation process is based on the occupancy probabilities of the grid cells and leads to the relaxation of the cubic grid cell assumption common to most 3D occupancy grids. The aim of this paper is to show the advantage of the proposed incremental fusion process which leads to the approximation of the 3D environment using rectangular cuboids. Evaluation on a large scale dataset and comparison to the state of the art shows that the proposed approach has faster access time for all occupied grid cells and requires a smaller number of cells for 3D environment representation.

The soft robotics approach is widely considered to enable robots in the near future to leave their cages and move freely in our modern homes and manufacturing sites. Musculoskeletal robots are such soft robots which feature passively compliant actuation, while leveraging the advantages of tendon-driven systems. Even though these robots have been intensively researched within the last decade, high-performance feedback control laws have only very recently been developed. In [1], a controller was developed utilizing Dynamic Surface Control (DSC), an extension to backstepping, with an adaptive neural network compensator for joint as well as muscle friction. We compare these novel control strategies to Computed Force Control (CFC), an existing technique from the field of tendon-driven control, yielding highly improved trajectory tracking. The musculoskeletal robot Anthrob [2] serves as a benchmark.

Need to develop human body's posture supervised robots, gave the push to researchers to think over dexterous design of exoskeleton robots. It requires to develop quantitative techniques to assess motor function and generate the command for the robots to act accordingly with complex human structure. In this paper, we present a new technique for the upper limb power exoskeleton robot in which load is gripped by the human subject and not by the robot while the robot assists. Main challenge is to find non-biological signal based human desired motion intention to assist as needed. For this purpose, we used newly developed Muscle Circumference Sensor (MCS) instead of electromyogram (EMG) sensors. MCS together with the force sensors is used to estimate the human interactive force from which desired human motion is extracted using adaptive Radial Basis Function Neural Network (RBFNN). Developed Upper limb power exoskeleton has seven degrees of freedom (DOF) in which five DOF are passive while two are active. Active joints include shoulder and elbow in Sagittal plane while abduction and adduction motion in shoulder joint is provided by the passive joints. To ensure high quality performance model reference based adaptive impedance controller is employed. Exoskeleton performance is evaluated experimentally by a neurologically intact subject which validates the effectiveness.

Activity recognition of multi-individuals (ARMI) within a group, which is essential to practical human-centered robotics applications such as childhood education, is a particularly challenging and previously not well studied problem. We present a novel adaptive human-centered (AdHuC) representation based on local spatio-temporal features (LST) to address ARMI in a sequence of 3D point clouds. Our human-centered detector constructs affiliation regions to associate LST features with humans by mining depth data and using a cascade of rejectors to localize humans in 3D space. Then, features are detected within each affiliation region, which avoids extracting irrelevant features from dynamic background clutter and addresses moving cameras on mobile robots. Our feature descriptor is able to adapt its support region to linear perspective view variations and encode multi-channel information (i.e., color and depth) to construct the final representation. Empirical studies validate that the AdHuC representation obtains promising performance on ARMI using a Meka humanoid robot to play multi-people Simon Says games. Experiments on benchmark datasets further demonstrate that our adaptive human-centered representation outperforms previous approaches for activity recognition from color-depth data.

This paper proposes a new approach for a group of robots carrying out a collaborative task to adapt on-line to actuation performance variations among the robots. We consider the problem of multi-robot coverage, where a group of robots has to spread out to cover the environment. We suppose that some robots have poor actuation performance (e.g. weak motors, friction losses in the gear train, wheel slip, etc.) and some have strong actuation performance (powerful motors, little friction, favorable terrain, etc.). The robots do not know before hand the relative strengths of their actuation compared to the others in the team. The algorithm in this paper learns the relative actuation performance variations among the robots on-line, in a distributed fashion, and automatically compensates by giving the weak robots a small portion of the environment, and giving the strong robots a larger portion. Using a Lyapunov-type proof, we prove that the robots converge to locally optimal positions for coverage. The algorithm is demonstrated in both Matlab simulations and experiments using Pololu m3pi robots.