IROS 2016 论文列表

This paper details a method for identifying a set of piece-wise affine linear models that can be used for control design for flapping-winged flight. The paper focuses on diving maneuvers as the application for these models. The flight conditions during the dive are segmented into separate dynamically similar regions, and least-squares is used to estimate affine linear models for each modeling region. These models are used to compute the reachability sets that satisfy recovery conditions for safe diving. The point within the dive to begin recovery was determined by checking the current pose for inclusion in the backward reachable set. Using this control method, 2.2 meter dives were achieved at a success rate of 60 percent. The data-driven automatic modeling techniques and controller design processes can be extended to additional flight maneuvers, provided sufficient previous data have been collected for model generation of those maneuvers.

We present FAST-Hex, a novel UAV concept which is able to smoothly change its configuration from underactuated to fully actuated by using only one additional motor that tilts all propellers at the same time. FAST-Hex can adapt to the task at hand by finely tuning its configuration from the efficient (but underactuated) flight (typical of coplanar multi-rotor platforms) to the full-pose-tracking (but less efficient) flight, which is attainable by non-coplanar multi-rotors. We also introduce a novel full-pose geometric controller for generic multi-rotors (not only the FAST-Hex) that outperforms classical inverse dynamics approaches. The controller receives as input any reference pose in ℝ3×SO(3) (3D position + 3D orientation). Exact tracking is achieved if the reference pose is feasible with respect to the propeller spinning rate saturations. In case of unfeasibility a new feasible desired trajectory is generated online giving priority to the positional part. The new controller is tested with the FAST-Hex but can be used for many other multi-rotor platforms: underactuated, slightly fully-actuated and completely fully-actuated.

Robots with Variable Stiffness Actuators (VSA) are intrinsically flexible in the joints. The built-in mechanical spring has the advantage of a higher peak performance, in some extend increased safety for humans interacting physically with the robot, and promises a more energy efficient robot for certain trajectories. This paper shows the modeling process of a VSA including energy losses on the example of the DLR Floating Spring Joint (FSJ). The model includes the full actuator dynamics with losses in electromechanical transformation of the motors. Furthermore, it models bearing and gear friction with stiction, Coulomb friction, viscous friction, and load dependent effects. With the obtained model the energy losses of benchmark trajectories are investigated and compared with a comparable stiff joint actuator.

In this paper, we address an issue on non-existence of a CAD model of an object in a model-based virtual visual servoing system. The approach for reconstructing a 3D model with fewer faces from point cloud data is presented. Because of a number of faces on the model surface, this reconstructed model required less rendering time than a CAD model, consequently, high speed and high accuracy model-based visual servoing can be achieved. We present the experimental results to demonstrate the effectiveness in terms of rendering time, convergence speed, and tracking speed measurements. The comparison of performances of a model-based visual servoing using different models, i.e.- CAD, Coarse mesh, Fine mesh, and Propose models, is also presented.

Micro-sized agents can benefit robotic minimally invasive surgery since they can be inserted into the human body and use natural pathways such as arteries and veins or the gastrointestinal tract, to reach their target for drug delivery or diagnosis. Recently, miniaturized agents with shape-changing and gripping capabilities have provided significant advantages in performing grasping, transportation, and manipulation tasks. In order to robustly perform such tasks, it is of utmost importance to properly estimate their overall configuration. This paper presents a novel solution to the problem of estimating and tracking the 3D position, orientation and configuration of the tips of miniaturized grippers from RGB marker-less visual observations obtained by a microscope. We consider this as an optimization problem, seeking for the gripper model parameters that minimize the discrepancy between hypothesized instances of the gripper model and actual observations of the miniaturized gripper. This optimization problem is solved using a variant of the Particle Swarm Optimization algorithm. The proposed approach has been evaluated on several image sequences showing the grippers moving, rotating, opening/closing and grasping biological material.

We report on the development of a dynamic gait in a robot based on a sagittal-plane and reduced-order model that fits the natural dynamics of the robot. The proposed model, referred to as the two-rolling-legs model with waist actuation (TRLW), consists of two rolling and compliant legs and two rigid bodies connected by an active waist joint. The numerical fixed-point analyses of the model suggest that with adequate touchdown conditions and waist motion, dynamic and periodic bounding and pronking behaviors of the model are achievable. Several identified fixed points of the model, or passive dynamics of the model, serve as the control guidance to initiate bounding and pronking on a newly-developed quadruped robot. The experimental results show that though several un-modeled factors cause behavioral discrepancies between the TRLW model and the robot, the robot can still initiate its dynamic behaviors by directly mapping the leg trajectories of the model onto those of the robot and via a simple pre-set joint position control strategy, without any other state feedback. This work confirms that the dynamic behavior of the robot can be initiated by using the passive dynamics of the reduced-order model.

A closed elastica, which is an elastic rod whose both ends are fixed to prescribed two places, can be found in various situations in robotics. In this paper, we investigate validity of the discretized spatial model of a closed elastica. The discretized closed elastica model can be regarded as a series of rigid bodies connected with passive elastic joints physically, and represented by a set of nonlinear equations derived by discretizing differential equations with boundary conditions as well as Euler equations expressing local minimum elastic energy. The validity of the considered model was studied by comparing the measured shape and the calculated shape from the model. Experimental results using a laser position sensor for shape measurement show that good correspondence between measured and calculated shapes, which supports the validity of the considered model. The constrained force and torque at the tip of an elastica, which can be obtained simultaneously associated with shape calculation, are also verified from measurement by a 6-axes force/torque sensor.

Spatial point process models are a commonly-used statistical tool for studying the distribution of objects of interest in a domain. We study the problem of deploying mobile robots as remote sensors to estimate the parameters of such a model, in particular the intensity parameter λ which measures the mean density of points in a Poisson point process. This problem requires covering an appropriately large section of the domain while avoiding the objects, which we treat as obstacles. We develop a control law that covers an expanding section of the domain and an online criterion for determining when to stop sampling, i.e., when the covered area is large enough to achieve a desired level of estimation accuracy, and illustrate the resulting system with numerical simulations.

This paper presents a mixed-integer model predictive controller for walking. In the proposed scheme, mixed-integer quadratic programs (MIQP) are solved online to simultaneously decide center of mass jerks, footsteps positions and steps durations while respecting actuation, geometry, and contact constraints. Simulation results show that this MIQP scheme is able to keep balance while a fixed step duration controller fails in situations where the robot faces very strong disturbances.

In this paper, we discussed the simultaneous intervention problem that arises from human-robot teams. The problem concerns with the case that one human operator has to intervene with several robots at almost the same time and it cannot be handled properly by existing methods. A model predictive control(MPC) based mixed initiative controller was proposed to solve this problem by embedding an intention model into the optimization problem. This method is in essence suitable for general MPC based mixed initiative controllers. The embedded intention model may cause the robot to deviate from its desired trajectory, for which conditions were developed to guarantee that the task completeness is not be affected. A comparison experiment was conducted and results showed that the proposed controller is more efficient and helpful in shortening the intervention time and allowing more robots to be successfully intervened in a simultaneous intervention scenario.

Animals are able to accelerate rapidly from rest with incredible dexterity but these transient motions are poorly understood. Here we present the first examination of the time optimal behaviour of a quadruped sprinting from rest. We develop a planar multi-body model and employ modern trajectory optimization methods to produce a motion without prescribing periodicity or foot contact order. Our trajectories produce several similarities with racing greyhounds. These results will inspire the design of feedback controllers for future manoeuvrable quadruped robots.

A mobile flying projector is hard to build due to the limited size and payload capability of a micro aerial vehicle. Few flying projector designs have been studied in recent research. However, to date, no practical solution has been presented. We propose a versatile laser projection system enabling in-flight projection with feedforward correction for stabilization of projected images. We present a quantitative evaluation of the accuracy of the projection stabilization in two autonomous flight experiments. While this approach is our first step towards a flying projector, we foresee interesting applications, such as providing on-site instructions in various human machine interaction scenarios.

This paper considers the merging of appearance-based spatial knowledge among robots having compatible visual sensing. Each robot is assumed to retain its knowledge in its individual long-term spatial memory where i) the place knowledge and their spatial relations are retained in an organized manner in place and map memories respectively; and ii) a ‘place’ refers to a spatial region as designated by a collection of associated appearances. In the proposed approach, each robot communicates with another robot, receives its memory and then merges the received knowledge with its own. The novelty of the merging process is that it is done in two stages: merging of place knowledge followed by the merging of map knowledge. As each robot's place memory is processed as a whole or in portions, the merging process scales easily with respect to the amount and overlap of the appearance data. Furthermore, the merging can be done in decentralized manner. Our experimental results with a team of three robots demonstrate that the resulting merged knowledge enables the robots to reason about learned places.

The paper focuses on the mechatronic design of a robotic manipulator that is meant to be mounted on an Unmanned Aerial Vehicle (UAV) and to be used in industrial applications, for both aerial inspection by contact and aerial manipulation. The combination of an UAV and the robotic manipulator realizes the aerial manipulator. The robotic manipulator is designed to be versatile so that the aerial manipulator can perform both trajectory tracking in free flight and physical interaction. Moreover, the robotic manipulator can be mounted on commercially available UAVs a modular way without interfering with the existing onboard control architecture. Experimental test are validating the overall mechatronic design.

This paper presents a novel method of stabilizing hybrid models of torque-constrained, underactuated walking robots - without using nonlinear gait optimization - by leveraging properties of the mechanics of the robot. At its core, the controller stabilizes the transfer of angular momentum from one leg to the next through continuous-time control coupled with hybrid system models that capture impacts that occur at foot strike. In particular, conservation of angular momentum at impact allows for computation of the exact transfer of momentum as a function of the robot's step length and vertical center of mass velocity just prior to foot impact. This motivates the construction of continuous-time reference trajectories for the robot's step length and vertical center of mass with endpoints corresponding to a desired transfer of angular momentum. Stabilization to these trajectories results in stable walking, as indicated by numeric Poincaré analysis. The controller is implemented in simulation of a five-link, underactuated 3D robot via Model Predictive Control which provides a means of achieving walking under non-trivial actuation limits.

This research illustrates the design, implementation, and evaluation of pneumatic variable-stiffness actuator (VSA) used to strum a four-stringed ukulele with audio variability. A guitar pick is antagonistically loaded with two inflatable polydimethylsiloxane (PDMS) actuators, allowing for the independent control of both position or stiffness through the utilization of one and two PDMS solenoids, respectively. To generate smooth analog pressure signals, the bellows incorporate a controlled leak to atmospheric pressure, having synonymous properties to a low-pass filter circuit when fed a coarse pressure signal through PWM control. Experimental results illustrate a minimum to maximum stiffness range of 76 to 320 Nmm/rad, a maximum cyclical speeds of 2.04 Hz, and a stiffness change rate of 96 Nmm/rad/kPa. Additional to the VSA, alternative PDMS fabrication methods are presented to allow for complex, precise manufacturing of silicone bodies in a low-cost manner.

We consider the question of benchmarking the performance of methods used for estimating the depth of a scene from a single image. We describe various measures that have been used in the past, discuss their limitations and demonstrate that each is deficient in one or more ways. We propose a new measure of performance for depth estimation that overcomes these deficiencies, and has a number of desirable properties. We show that in various cases of interest the new measure enables visualisation of the performance of a method that is otherwise obfuscated by existing metrics. Our proposed method is capable of illuminating the relative performance of different algorithms on different kinds of data, such as the difference in efficacy of a method when estimating the depth of the ground plane versus estimating the depth of other generic scene structure. We showcase the method by comparing a number of existing single-view methods against each other and against more traditional depth estimation methods such as binocular stereo.

Providing guarantees on the system behavior is mandatory in order to let the robots enter our every-day life. Among these guarantees, proving the fulfillment of real-time constraints on the software is a key issue, as their violation could result into unexpected and unsafe behaviors. In this paper, we present a methodology to guarantee real-time constraints on component-based software architectures of robots. This methodology relies on the MAUVE language to model the component architecture, and on a set of analysis tools that first estimate the worst case execution time of elementary functions from actual component traces, and then check the real-time constraints of each component. We illustrate this process on the architecture developed for the autonomous navigation of a partially known area by a mobile robot.

In order to facilitate and improve robots social acceptance, they must be equipped with behaviors similar to those of humans. It is therefore necessary to study and model the phenomenon to be reproduce. This paper studies and analyzes the physical parameters of the handshake in order to have its characteristic features (frequency, duration, strength, synchronization, etc.) used to model this interaction. Features that would later help to develop bio-inspired adaptive controllers, which will allow humanoid robots to better interact with humans according to simple social contexts.