IROS 2016 Papers

In dynamic environments crowded with people, robot motion planning becomes difficult due to the complex and tightly-coupled interactions between agents. Trajectory planning methods, supported by models of typical human behavior and personal space, often produce reasonable behavior. However, they do not account for the future closed-loop interactions of other agents with the trajectory being constructed. As a consequence, the trajectories are unable to anticipate cooperative interactions (such as a human yielding), or adverse interactions (such as the robot blocking the way). In this paper, we propose a new method for navigation amongst pedestrians in which the trajectory of the robot is not explicitly planned, but instead, a planning process selects one of a set of closed-loop behaviors whose utility can be predicted through forward simulation. In particular, we extend Multi-Policy Decision Making (MPDM) [1] to this domain using the closed-loop behaviors Go-Solo, Follow-other, and Stop. By dynamically switching between these policies, we show that we can improve the performance of the robot as measured by utility functions that reward task completion and penalize inconvenience to other agents. Our evaluation includes extensive results in simulation and real-world experiments.

Complexity, cost, and power requirements for actuation of individual robots are large factors in limiting the size of robotic swarms. Here we present a prototype robotic system that allows for externally powered motion in 2D without sacrificing individual autonomy, which simplifies the robot hardware, possibly enabling larger swarm sizes. This is accomplished using a table surface that is moving in an orbital fashion, and where robots can move to any point on the table surface simply through a series of carefully timed attachment and detachment steps. We present a model for the robot's motion, and use this model to create a motion controller that allows the robot to move from its current position to any other position on the table in approximately a straight line. We show this controller working in simulation as well as on an experimental hardware system.

A labeled motion library, in which robot motions are associated with semantic meanings, e.g., words, is useful for human-robot interaction, as a robot can use it to autonomously select motions to support its non-verbal communication. Manually assigning labels to new motions to a motion library is time consuming. However, a new motion may be similar to motions in the labeled motion library, and can be mapped to existing labels. We formally define motions, labels, and mappings between motions and labels. We use a NAO humanoid robot as a motivating example, though our approach is general for use on a humanoid robot with rotational joints. We explain how we generate motions and labels, define eight distance metrics to determine the similarity between motions, and use the nearest neighbor algorithm to determine the labels of a new motion. The distance metrics are varied across three axes - Euclidean versus Hausdorff, joint angles versus points of interest (postures), and mirrored versus non-mirrored. We evaluate the efficacy of these eight distance metrics, using precision, recall, and computational complexity.

Policy Gradient methods require many real-world trials. Some of the trials may endanger the robot system and cause its rapid wear. Therefore, a safe or at least gentle-to-wear exploration is a desired property. We incorporate bounds on the probability of unwanted trials into the recent Contextual Relative Entropy Policy Search method. The proposed algorithm is evaluated on the task of autonomous flipper control for a real Search and Rescue rover platform.

In this paper, we design an autonomous flight controller for height regulation and bang-bang controller for directional control of a light-weight flapping-wing micro air vehicle (FWMAV) with limited payload. We also present autonomous vision-based target tracking for a FWMAV equipped with a low-cost and light-weight first person view (FPV) camera. We construct a ground station, integrated with control and vision algorithms, which performs the image processing and the computation of control inputs based on the acquired state variables from motion capture system. In addition, we employ a vision algorithm for a low-quality camera to detect a static target with the discussions on the techniques to improve the reliability of visual detection. Experimental results show satisfactory flight performance, achieving the height regulation and directional control, and autonomous vision-based target tracking.

We present a novel fault detection method for application in component-based robotic systems. In contrast to existing work, our method specifically addresses faults in the software system of the robot using a data-driven methodology which exploits the inter-process communication of the system. This enables an application of the approach without expert knowledge or availability of complex software models. We specifically focus on performance bugs, which slowly degrade the performance of the system and are thereby harder to detect but also most valuable for automatic recovery. Using a data set recorded on a RoboCup@Home platform we demonstrate the performance and applicability of our method and analyze the types of faults that can be detected by the method.

Building robots from soft materials provides opportunities to create more robust and adaptive designs: soft bodies can conform to complex shapes in the environment and they are able to cushion shocks and store elastic energy. The challenge, however, remains to control highly deformable moveable structures effectively. We have proposed that one useful approach is through an autonomous decentralized control using physically coupled mechano-sensory oscillators. We have developed a highly deformable 3-D printed soft robot (PS robot) as a platform to explore the validity of the control strategy. Based on this platform we introduce a caterpillar-like soft-bodied modular-robot that is controlled in a fully decentralized manner. This paper focuses on one of the advantageous characteristics of autonomous decentralized control, i.e., the extensibility/contractibility of the modularity structures. The numerical and experimental results demonstrate that simple oscillators (controllers) can interact with each other by deforming the soft-bodied structure leading to a phase gradient that propagates crawling locomotion. The PS robots provide a powerful platform for exploring this strategy and they can be generalized for use with different robot shapes and material properties.

Aggregation of self-reconfigurable robotic modules can potentially offer many advantages for robotic locomotion and manipulation. The resulting system could be more reliable and fault-tolerant and provide the necessary flexibility for new tasks and environments. However, self-aggregation of modules is a challenging task, especially when the alignment of the docking parties in a 3D environment involves both position and orientation (6D), since the bases of docking may be non-stationary (e.g., floating in space, underwater, or moving along the ground), and the end-effectors may have accumulated uncertainties due to many dynamically-established connections between modules. This paper presents a new framework for docking in such a context and describes a solution for sensor-guided self-reconfiguration and manipulation with non-fixed bases. The main contributions of the paper include a realistic experiment setting for 6D docking where a modular manipulator is floating or rotating in space with a reaction wheel and searches and docks with a target module using vision. The movement of the docking parties is a combination of floating and manipulation, and the precision of the docking is guided by a sensor located at the tip of the docking interface. The docking itself is planned and executed by a real-time algorithm with a theoretical convergence boundary. This new framework has been tested in a high-fidelity physics-based simulator, as well as by real robotic modules based on SuperBot. Experimental results have shown an average success rate of more than 86.7 percent in a variety of different 6D-docking scenarios.

The precise location of tumors is an important step in surgical planning that can be obtained from mechanical properties of soft tissues. In this paper we propose a robotic-assisted palpation system that automatically moves an ultrasound probe to optimize the elastography process and improve the resulting elastogram. The main contribution of this work is the use of the elastography modality directly as input of the robot controller. Force measures are also considered in the probe control in order to automatically induce soft tissue deformation needed for real-time elastography imaging process. Moreover, an automatic exploration process is implemented to orient the probe to reach different views of a soft tissue target of interest. This allows to improve the elastogram quality of the element of interest by fusing the information observed from different positions.

Ultrasound (US) imaging is commonly employed for the diagnosis and staging of abdominal aortic aneurysms (AAA), mainly due to its non-invasiveness and high availability. High inter-operator variability and a lack of repeatability of current US image acquisition impair the implementation of extensive screening programs for affected patient populations. However, this opens the way to a possible automation of the procedure, and recent works have exploited the use of robotic platforms for US applications, both in diagnostic and interventional scenarios. In this work, we propose a system for autonomous robotic US acquisitions aimed at the quantitative assessment of patients' vessel diameter for abdominal aortic aneurysm screening. Using a probabilistic measure of the US quality, we introduce an automatic estimation of the optimal pressure to be applied during the acquisition, and an online optimization of the out-of-plane rotation of the US probe to maximize the visibility of the aorta. We evaluate our method on healthy volunteers and compare the results to manual acquisitions performed by a clinical expert, demonstrating the feasibility of the presented system for AAA screening.

Vehicle performance tests are conducted to evaluate factors of vehicle performance such as fuel consumption and durability. Robotic drivers are often used in these tests to ensure that performance evaluate is reproducible, and various types of vehicles have been tested. Manual transmission (MT) vehicles are widely used and represent about 50% of vehicles in the world. When the driver selects the gear ratio in MT, speed control performance is degraded during changing the gear. Because a torque transmission between the engine and wheels is cut during this sequence, the driver must drive the vehicle while considering the influence of changing the gear. Therefore, control performance in MTs depends on driver's technique. If robotic drivers are used with MT, there is a possibility that performance will be significantly degraded because the robot cannot adjust to change the gear while driving. In this paper, we realized the driving while considering shift changes by the robotic driver. First, we modeled the lack of a driving force using the dynamic characteristics of the vehicle and the deceleration during changing the gear. We then derived a target vehicle speed waveform using the lack of a driving force model. Finally, we performed vehicle tests using a robotic driver. We confirmed the effectiveness of the proposed system from the results of actual vehicle tests.

The performance of a ROS application is a function of the individual performance of its constituent nodes. Since ROS nodes are typically configurable (parameterised), the specific parameter values adopted will determine the level of performance generated. In addition, ROS applications may be distributed across multiple computation devices, thus providing different options for node allocation. We address two configuration problems that the typical ROS user is confronted with: i) Determining parameter values and node allocations for maximising performance; ii) Determining node allocations for minimising hardware resources that can guarantee the desired performance. We formalise these problems with a mathematical model, a constrained form of a multiple-choice multiple knapsack problem. We propose a greedy algorithm for optimising each problem, using linear regression for predicting the performance of an individual ROS node over a continuum set of parameter combinations. We evaluate the algorithms through simulation and we validate them in a real ROS scenario, showing that the expected performance levels only deviate from the real measurements by an average of 2.5%.

State-of-the-art neural microprobes contain hundreds of electrodes within a single shaft. Due to hardware and wiring restrictions, it is usually only possible to measure a small subset of the available electrodes simultaneously. The selection of the best channels is typically performed offline either manually or automatically. However, having a fixed selection for long-term observation does not allow the system to react to changes in the neural activity, and may therefore lead to the loss of important information. In this paper, we formulate the process of autonomously selecting the best subset of electrodes as a combinatorial multi-armed bandit problem with non-stationary rewards, thus allowing the probe to adapt its selection policies online. In order to minimize exploratory actions of the probe, we furthermore take advantage of the existing dependencies between neighboring channels. Our approach is an adaptation of the discounted upper confidence bounds (D-UCB) algorithm, and identifies the electrodes providing the largest amount of non-redundant information. To the best of our knowledge, this is the first online approach for the problem of electrode selection. In extensive experiments, we demonstrate that our solution is not only able to converge towards an average optimal selection policy, but it is also able to react to changes in the neural activity or to damages of the recording electrodes.

In a hierarchical motion planning system for urban autonomous driving, it is a common practice to separate tactical reasoning from the lower-level trajectory planning. This separation makes it difficult to achieve robust maneuver-based tactical reasoning, which is intrinsically linked to trajectory planning. We therefore propose a planning method that automatically discovers tactical maneuver patterns, and fuses pattern reasoning and sampling-based trajectory planning. The results demonstrate enhanced planning feasibility, coherency and scalability.

It is of great importance to pick up a single carbon nanotube (CNT) from a bulk of CNTs for nanodevice fabrication. In this study, we have proposed a nanorobotic manipulation system allowing automated pick-up of CNTs based on visual feedback. We utilize histogram normalization for automatic binarization, and it achieves to clearly distinguish CNTs from substrate and other impurities under different image brightness. Furthermore, we develop the gradient orientation inversion (GOI) algorithm to recognize CNT tip and atomic force microscopy (AFM) cantilever. Taking full advantages of the geometrical characteristics of CNT and AFM cantilever, GOI is proved to be quite robust. We have designed segment detection method (SDM) to successfully separate the AFM cantilever and CNT, whereas the contact detection between them is achieved by analyzing the straightness variation. Preliminary experimental results imply that our method shows high promise in realistic fabrication of nanodevices.

As rapid development of precision medicine, in vivo manipulation of micro/nano-scaled particles have attracted increasing attention in recent years. To accommodate complex in-vivo environment, robot-aided automated manipulation technology is highly demanded in trapping and controlling micro/nano-particles stably and effectively. This paper presents an in-vivo cell manipulation system, where a disturbance compensation controller is utilized to minimize the effect of fluid (e.g., blood flow) on the cell. The controller has exhibited advantages in adjusting cell tracking trajectory online, minimizing the steady-state error, and eliminating overshoot. Simulation and experimental results verify the performance of the controller.

Water is a highly abundant nutrient in the human body and monitoring of its regulation is essential to keep the body hydrated. A number of critical health conditions including swelling of the brain and short/long term memory loss are associated with poor or excessive drinking habits. This can be prevented with the use of a real time hydration monitoring system. In this paper we presented AutoHydrate, a wearable hydration monitoring system which continuously monitors the drinking activities and daily fluid requirements of the user through automatic detection of drinking and body activities. The system is built using a throat microphone for collecting acoustic signals, a smartwatch for collecting body activity, an embedded computer for processing the signals and sending recommendation to a smartphone app in real time for an interactive information display. After different time, frequency and cepstral domain features are extracted from the signals, drinking activities are classified using Support Vector Machine (SVM) and body activity is classified using Gradient Boosting Decision Tree algorithm. The Dietary Reference Intake standard is followed for recommending the amount of fluid required using our detection. Based on our experimental results on 8 subjects, a Drinking detection accuracy of 91.5% and Body activity classification accuracy of 89.12% are obtained. Results show that our system is feasible for real time monitoring of body hydration.

Deep convolutional neural networks are being increasingly deployed for image classification tasks as they can learn sensor and environmental independence from large quantities of training data. Most, however, have focused on classifying uploaded photographs rather than the often occluded, arbitrary height and camera angles images found commonly in robotic applications. In this work, we look at the performance of the popular AlexNet architecture to detect people in different robotic scenarios using different sensors and/or environments. Furthermore, we demonstrate how fusing this architecture with the depth-based layered detection system, a more traditional geometric feature-based classifier, leads to significant improvements in classification precision/recall, whether working with depth data alone or a combination of depth and RGB images.

In this paper we propose a novel control approach for robots that are physically coupled with humans, such as exoskeletons and assisting devices. In contrast with the conventional controllers, where assistance is usually provided indifferently of the configuration of the arm and direction of the motion, we propose a control method that compensates the anisotropic property of the manipulability related to the human arm. Consequently, the assistive behaviour of the proposed method allows the user to perform tasks in arm configurations that are otherwise unsuitable due to the lack of manipulability. In effect, the proposed method transforms the elliptic shape of the force manipulability ellipse to a circular shape in the whole space of human arm configurations. The proposed approach was evaluated by the arm-reaching task that involved pushing of a heavy object on a plane. We tested the accuracy and efficiency of the proposed method under several conditions that involved motions with added weight and friction. The results of our study show that the proposed approach significantly improves the human motor control ability and maintains the desired accuracy of the movement.