IROS 2018 论文列表

Robots operating in human environments are often required to recognise, grasp and manipulate objects. Identifying the locations of objects amongst their complex surroundings is therefore an important capability. However, when environments are unstructured and cluttered, as is typical for indoor human environments, reliable and accurate object segmentation is not always possible because the scene representation is often incomplete or ambiguous. We overcome the limitations of static object segmentation by enabling a robot to directly interact with the scene with non-prehensile actions. Our method does not rely on object models to infer object existence. Rather, interaction induces scene motion and this provides an additional clue for associating observed parts to the same object. We use a probabilistic segmentation framework in order to identify segmentation uncertainty. This uncertainty is then used to guide a robot while it manipulates the scene. Our probabilistic segmentation approach recursively updates the segmentation given the motion cues and the segmentation is monitored during interaction, thus providing online feedback. Experiments performed with RGB-D data show that the additional source of information from motion enables more certain object segmentation that was otherwise ambiguous. We then show that our interaction approach based on segmentation uncertainty maintains higher quality segmentation than competing methods with increasing clutter.

This paper presents an underwater robotic sensor system for localizing acoustic transmitters when the robot's hydrophones cannot be time-synchronized. The development of the system is motivated by applications where tracking of marine animals that are tagged with an underwater acoustic transmitter is required. The system uses two novel real-time calibration algorithms that improve the accuracy of time of flight (TOF) and time difference of arrival (TDOA) measurements. The first algorithm corrects non-linear clock skews in TOF measurements based on temperature variation. The second algorithm compensates the localized relative clock skew between clocks using a mixed integer linear program. To validate the system's performance, an Autonomous Underwater Vehicle (AUV) was deployed to track a moving tag where GPS data was used as ground truth. Compared to traditional TOF and TDOA filtering methods, the results show that the proposed system can achieve reduction of mean localization errors by 59%, and a reduction of the standard deviation of measurements by 44%.

We propose a novel sensing method for soft pneumatic actuators. The method uses a single microphone, embedded into the actuator's air chamber. Contact with the environment induces sound (vibration) in the actuator. The materials and the shape of the actuator reflect, refract, and attenuate the sound as it propagates inside the actuator. This produces a unique sound signature for different types of events, enabling the sensing of contact locations, contact force, and the type of contacted material. Sensing is insensitive to the inflation state of the actuator and to background noise. We demonstrate the robustness and versatility of the microphone-based sensor solution in experiments with a PneuFlex actuator. The proposed sensorization avoids the fundamental challenges of sensorizing soft pneumatic actuators, because the placement of a microphone does not negatively affect the compliance of the actuator and because a single microphone suffices for sensorization of the entire actuator, eliminating the need for an application-specific sensor layout.

Prefabrication of structures is currently used in a limited capacity, due to the lack of flexibility, despite the potential cost and speed advantages. Autonomous flexible reassembly enables structures to be developed which can be continuously and iteratively dis-assembled and re-assembled providing far more flexibility in comparison to single shot pre-fabrication methods. Dis-assembly of structures should be considered when assembling, due to the asymmetry of assembly and dis-assembly processes, to ensure structures can be recycled and re-assembled. This allows for agile development, significantly reducing the time and resource usage during the build process. In this work, a framework for flexible re-assembly is developed and a robotic platform is developed to implement and test this framework with simple Lego bricks. The tradeoffs in terms of time, resource use and probability of success of this new assembly method can be understood by using a cost function to compare to alternative fabrication methods.

Highly accurate mapping and localization is of prime importance for mobile robotics, and its core lies in efficient scan matching. Previous research are focusing on designing a robust objective function and the residual error distribution is often ignored or simply assumed as unitary or mixture of simple distributions. In this paper, a mixture of exponential power (MoEP) distributions is proposed to approximate the residual error distribution. The objective function induced by MoEP-based residual error modelling ensembles a mix-norm-based scan matching (MiNoM), which enhances the matching accuracy and convergence characteristic. Both the parameters of transformation (rotation and translation) and residual error distribution are estimated efficiently via an EM-like algorithm. The optimization of MiNoM is iteratively achieved via two phases: An on-line parameter learning (OPL) phase to learn residual error distribution for better representation according to the likelihood field model (LFM), and an iteratively reweighted least squares (IRLS) phase to attain transformation for accuracy and efficiency. Extensive experimental results validate that the proposed MiNoM out-performs several state-of-the-art scan matching algorithms in both convergence characteristic and matching accuracy.

Deformable objects like cloth and rope are challenging to manipulate because it is difficult to predict the state of the object given a motion of the gripper(s) holding it. In much previous work, physical models (such as Mass-Spring or Finite-Element) have been used to model such affects. However, these models often require significant parameter tuning for each scenario and can be expensive to simulate inside a control loop. Furthermore, it is difficult to create a practical controller for deformable object manipulation that preserves constraints, especially avoiding overstretching the object. In this paper, we developed a more effective controller than previous work by (1) constructing a more accurate geometric model of how the direction of gripper motion and obstacles affect deformable objects; and (2) specifying a novel stretching avoidance constraint to prevent the object from being overstretched by the robot. Experiments comparing our new method to the previous method in simulation and on a physical robot suggest that our new model captures the behavior of the object more accurately. We also find that our controller is able to prevent tearing that would occur when using the previous method.

We propose using the Successor Representation (SR) to accelerate learning in a constructive knowledge system based on General Value Functions (GVFs). In real-world settings, like robotics for unstructured and dynamic environments, it is impossible to model all meaningful aspects of a system and its environment by hand. Instead, robots must learn and adapt to changes in their environment and task, incrementally constructing models from their own experience. GVFs, taken from the field of reinforcement learning (RL), are a way of modeling the world as predictive questions. One approach to such models proposes a massive network of interconnected and interdependent GVFs, which are incrementally added over time. It is reasonable to expect that new, incrementally added predictions can be learned more swiftly if the learning process leverages knowledge gained from past experience. The SR provides a means of capturing regularities that can be reused across multiple GVFs by separating the dynamics of the world from the prediction targets. As a primary contribution of this work, we show that using the SR can improve sample efficiency and learning speed of GVFs in a continual learning setting where new predictions are incrementally added and learned over time. We analyze our approach in a grid-world and then demonstrate its potential on data from a physical robot arm.

We propose a hybrid approach aimed at improving the sample efficiency in goal-directed reinforcement learning. We do this via a two-step mechanism where firstly, we approximate a model from Model-Free reinforcement learning. Then, we leverage this approximate model along with a notion of reachability using Mean First Passage Times to perform Model-Based reinforcement learning. Built on such a novel observation, we design two new algorithms - Mean First Passage Time based Q-Learning (MFPT-Q)and Mean First Passage Time based DYNA (MFPT-DYNA), that have been fundamentally modified from the state-of-the-art reinforcement learning techniques. Preliminary results have shown that our hybrid approaches converge with much fewer iterations than their corresponding state-of-the-art counterparts and therefore requiring much fewer samples and much fewer training trials to converge.

Acting systems aim at refining high-level actions into executable commands, while managing access to resources, possible failures, or any other unpredictable situation. Improving the trust on autonomous robots also requires to have a formal model of acting, and the capability to perform some analysis on this model. In this paper, we present ASPiC, an acting system based on the modeling of robot's skills using a specific control-flow Petri net model. The skills can then be combined using well-defined operators to build a complete plan that refines a high-level action. Some properties are guaranteed by construction, while others can be verified on the resulting plan model. ASPiC is finally applied to an area protection mission by an autonomous surface vehicle.

Precise Point Positioning (PPP) offers advantages over Real-Time Kinematic (RTK) positioning in that it needs no ground base station or a rover to provide a communication channel with the base station. Therefore, PPP is expected to be applied in various fields, such as precise agriculture, intelligent transportation, construction, and mining. One major drawback of PPP is that it takes several tens of minutes for initial convergence to reach an adequate level of accuracy, and how to shorten the convergence time remains the key issue regarding the proliferation of PPP. In our previous work, we proposed a method of drastically reducing this initial convergence time of PPP (called “ARIADNE”) by using an accurate fiducial marker whose position in Earth-fixed coordinates has been accurately measured using an onboard camera. In this contribution, in order to improve both the positioning accuracy and reliability, we introduce some new techniques to ARIADNE: (1) carrier phase ambiguity resolution (AR); and (2) estimating displacement of the marker's orientation. AR results in doubling the positioning accuracy, while displacement estimation enables the detection of any change in the marker's installation orientation and compensation for the effect thereof in the initialization process. Analytical results based on actual data acquired with a prototype system show that the above method and techniques work very well with a realistic setup of PPP-AR and marker performance, and successfully reduce the initial convergence time from tens of minutes to less than a minute.

In 1999 Sony released the AlBO Entertainment Robot, selling more than 150,000 units worldwide until 2006. By 2014, Sony had stopped offering upgrades and maintenance for the product, and owners were faced with the fact their pet-like robots would “die”. Some shrines and temples in Japan hold ningyo kuyō or mass funerals for dolls and other toys. At the suggestion of a small Japanese tech-repair company called A-Fun, one temple began offering a Buddhist funeral ceremony for AIBOs. Approximately 700 AIBOs have so far received a funeral service. This paper surveys A-Fun‘s maintenance services for old AIBOs, the AIBO funerals, and Sony's new 2018 AIBO release, in the cross-disciplinary context of human-machine relations in Japan and elsewhere. Drawing on the author's interviews with key actors, it articulates links between philosophy and neuroscience to explain tendencies toward zoomorphism in robot design. Perceiving presence (sonzai kan) and sensibility (kansei) in objects is a culturally contingent phenomenon. Whereas ways of conceiving the partly animate are largely absent from Western philosophy, in the case of AIBO ownership in Japan there is a reverential mindfulness of the technology's inherent contradictions.

This paper presents a method for reconstructing globally consistent 3D High-Definition (HD) maps at city scale. Current approaches for eliminating cumulative drift are mainly based on the pose graph optimization under the constraint of scan-matching factors. The misaligned edges in the graph may have negative impacts on the results. To address this problem and further handle inconsistency caused by multi-task acquisitions in urban environments, we introduce a refined structure of the factor graph considering systematical initialization bias, where the scan-matching factors are twice validated through a novel classifier and a robust optimization strategy. In addition, we incorporate a multi-hypothesis extended Kalman filter (MH-EKF) to remove dynamic objects. Quantitative experimental results demonstrate that the proposed method outperforms state-of-the-art techniques in terms of map quality.

We present in this paper a modular real-time model-based visual tracker. It is able to fuse different types of measurement, that is, edge points, textured points, and depth map, provided by one or multiple vision sensors. A confidence index is also proposed for determining if the outputs of the tracker are reliable or not. As expected, experimental results show that the more various measurements are combined, the more accurate and robust is the tracker. The corresponding C++ source code is available for the community in the ViSP library.

This paper presents a (patent-pending) small, quasi-static, minimal-complexity Stair Climbing Robot (SCR). The vehicle design is given simply by adding a third motor to a (Segway-like) Mobile Inverted Pendulum (MIP), enabling it to maneuver up stairs, leveraging feedback control, by planting it's “foot” onto the ground in front of the next step, lifting the chassis/wheel assembly up it's own “leg”, leaning over onto the top of the next step, self uprighting, and repeating for the following step(s). Fore/aft stabilization during leg balancing is given by using the MIP drive wheels as reaction wheels, while left/right stability is given by the width of the foot itself. The design is small and simple enough to potentially be ruggedized as a stair-climbing throwbot, akin to the Recon Scout (but able to climb up stairs) for reconnaissance in military and homeland security applications.

We present a distributed model for a team of autonomous aerial robots to collaboratively track a target without external control. The model uses distributed consensus to coordinate actions and to maintain formation via geometric constraints. Each robot uses its ego-centric view of a target and the relative distance from its two closest neighbors to infer its steering commands. To account for noisy and missing target detections, the robots exchange their estimated target position and formation configuration through shared PID-controlled steering responses. We show that the proposed model enables the team to maintain the view of a maneuvering target with varying acceleration under noisy detections and failures up to situations when all robots but one lose the target from their field of view.

Robots generate large amounts of data which need to be stored in a meaningful way such that they can be used and interpreted later. Such data can be written into log files, but these files lack the querying features and scaling capabilities of modern databases - especially when dealing with multi-robot systems, where the trade-off between availability and consistency has to be resolved. However, there is a plethora of existing databases, each with its own set of features, but none designed with robotic use cases in mind. This work presents three main contributions: (a) structures for benchmarking scenarios with a focus on networked multi-robot architectures, (b) an extensible workbench for benchmarking databases for different scenarios that makes use of Docker containers and (c) a comparison of existing databases given a set of multi-robot use cases to showcase the usage of the framework. The comparison gives indications for choosing an appropriate database.

This paper presents a feature encoding method of complex 3D objects for high-level semantic features. Recent approaches to object recognition methods become important for semantic simultaneous localization and mapping (SLAM). However, there is a lack of consideration of the probabilistic observation model for 3D objects, as the shape of a 3D object basically follows a complex probability distribution. Furthermore, since the mobile robot equipped with a range sensor observes only a single view, much information of the object shape is discarded. These limitations are the major obstacles to semantic SLAM and view-independent loop closure using 3D object shapes as features. In order to enable the numerical analysis for the Bayesian inference, we approximate the true observation model of 3D objects to tractable distributions. Since the observation likelihood can be obtained from the generative model, we formulate the true generative model for 3D object with the Bayesian networks. To capture these complex distributions, we apply a variational auto-encoder. To analyze the approximated distributions and encoded features, we perform classification with maximum likelihood estimation and shape retrieval.

This paper presents a soft actuator embedded with conductive liquid metal and shape memory epoxy (SME) which function together to enable self-sensing, tunable mechanical degrees of freedom (DoF), and variable stiffness. We embedded thermoplastic shape memory epoxy in the bottom portion of the actuator. Different sections of the SME could be selectively softened by an implanted conductive silver yarn located at different positions. When an electric current passes through the conductive silver yarn, it induces a phase transition that changes the epoxy from stiff state to compliant state. Each section of SME could be softened within 5 s by applying a current of 200 mA to the silver yarn. To acquire the strain curvature, eGaIn was infused into a microchannel surrounding the chambers of the soft actuator. A spiral-shaped eGaIn sensor was also attached to the tip of the actuator to perceive the contact with reliable dynamic force response. Systematic experiments were performed to characterize the stiffness, tunable DoF, and sensing property. We show the ability of the soft composite actuator to support a weight of 200g at the tip (as a cantilever) while maintaining the shape and the ability to recover its original shape after large bending deformation. In particular, seven different motion patterns could be achieved under the same pneumatic pressure of the actuator due to selectively heating the SME sections. A gripper which was fabricated by assembling two actuators to a base was able to grasp the weight up to 56 times of a single actuator through an appropriate motion pattern. For demonstration purposes, the gripper was used to grasp various objects by adjusting the DoF and stiffness with real-time feedback of the bending strain and the contact force.

The universal gripper has attracted attention due to its simple structure and advanced grasping ability for irregularly shaped objects. In this research, we propose a novel design for a granular-jamming-based gripper which uses a transparent filling and a semi-transparent membrane to allow optical sensing to detect both deformation of the membrane and the object being grasped. By adjusting the refractive index of an oil mixture to the refractive index of the granular bodies, we produced a fully transparent filling that allows the use of a camera inside the universal gripper. In this paper, we present the materials and development of our prototype, and describe the experimental confirmation of the prototype's performance. We showed that our prototype was able to grasp cylindrical and rectangular objects between 10 to 70 mm length while also tracking the deformation of the gripper.