ICRA 2017 Papers

In this paper, we present an improved octree-based mapping framework for autonomous navigation of mobile robots. Octree is best known for its memory efficiency for representing large-scale environments. However, existing implementations, including the state-of-the-art OctoMap [1], are computationally too expensive for online applications that require frequent map updates and inquiries. Utilizing the sparse nature of the environment, we propose a ray tracing method with early termination for efficient probabilistic map update. We also propose a divide-and-conquer volume occupancy inquiry method which serves as the core operation for generation of free-space configurations for optimization-based trajectory generation. We experimentally demonstrate that our method maintains the same storage advantage of the original OctoMap, but being computationally more efficient for map update and occupancy inquiry. Finally, by integrating the proposed map structure in a complete navigation pipeline, we show autonomous quadrotor flight through complex environments.

This paper presents the ongoing work towards enabling robots to achieve highly dynamic behavior through full-body teleoperation. Human operator and robot slave have independent balance controllers that interact with each other during the experiments. First we present a compliant balancing controller that regulates the feet contact forces in order to mitigate external disturbances and maintain balance. Next, by estimating the forces that the operator exerts over its own Center of Mass to generate movement, the Dynamic Synchronization Force Scaling controller allows the robot to anticipate human motion during posture tracking. This strategy requires reduced control gains for state tracking when compared to purely reactive controllers, resulting in an inherently more stable system. Results show a considerable reduction of the position tracking overshoot along with substantial reduction of required error-based control forces.

High contour accuracy is vital to the multi-axis motion system. Improvement in the contour accuracy of the parallel kinematic machine (PKM) has being a challenging issue in the process of its practical application. In analogy to the intelligent structure of the organisms, this paper proposes a smart structure based compensation (SSBC) method to improve the PKM's contour accuracy. Instead of adopting the advanced control strategies or improving the performances of machine's hardware, this method deals with the contouring issue through innovative mechanical design. The smart structures, which have the ability of perception and control, are introduced into the PKM's components to actively suppress their deformations and vibrations. Meanwhile, the accurate micro motions provided by the smart structure components can be used to compensate for the PKM's contour error caused by the uncoordinated control of the driving system just by using a simple control strategy. The implementation procedure of the SSBC method is first discussed in detail. A 2-DOF PKM is taken as the application object and the smart structures are introduced into its kinematic chains to control their axial deformations and internal forces. Then this PKM's kinematic error is reduced in advance to guarantee the result of the contour accuracy improvement. Finally, several experimental tests are performed to verify the effectiveness the proposed SSBC method in improving the PKM's contour accuracy. Except for the PKM, the SSBC method also provides a novel and effective way for the advanced multi-axis machines to improve their contour accuracies.

Conventional approaches to semantic segmentation are inappropriate for robotic applications, as they focus on pixel-level performance and give little significance to spurious object detections. This paper presents a region-based conditional random field model for semantic segmentation that focuses on object-level performance, recognising that in a robotics context, false object detections can have costly consequences. We show how optimising at the semantic region-level results in significantly fewer false positive object detections than conventional approaches. We further show how both object and pixel-level performance can be improved over conventional methods by combining region random fields with dense pixel random fields in a hierarchical manner. An object-aware performance metric is introduced that heavily penalises false positive and false negative object detections, as appropriate for robotic applications. Our approach is evaluated on the challenging NYU v2 and Pascal VOC datasets, outperforming comparable conventional methods in terms of object and pixel-level performance.

Traditional robotic surgical systems rely entirely on robotic arms to triangulate articulated instruments inside the human anatomy. This configuration can be ill-suited for working in tight spaces or during single access approaches, where little to no triangulation between the instrument shafts is possible. The control of these instruments is further obstructed by ergonomic issues: The presence of motion scaling imposes the use of clutching mechanics to avoid the workspace limitations of master devices, and forces the user to choose between slow, precise movements, or fast, less accurate ones. This paper presents a bi-manual system using novel self-triangulating 6-degrees-of-freedom (DoF) tools through a flexible elbow, which are mounted on robotic arms. The control scheme for the resulting 9-DoF system is detailed, with particular emphasis placed on retaining maximum dexterity close to joint limits. Furthermore, this paper introduces the concept of gaze-assisted adaptive motion scaling. By combining eye tracking with hand motion and instrument information, the system is capable of inferring the user's destination and modifying the motion scaling accordingly. This safe, novel approach allows the user to quickly reach distant locations while retaining full precision for delicate manoeuvres. The performance and usability of this adaptive motion scaling is evaluated in a user study, showing a clear improvement in task completion speed and in the reduction of the need for clutching.

This paper presents a novel caterpillar inspired rolling gait generation and control mechanism of a spherical robot. The research investigates efficient locomotory rolling gaits of Pleurotya caterpillar in unstructured environment. A similar rhythmic rolling pattern is produced for the spherical robot locomotion. A synergetically combined feedforward - feedback control strategy is further proposed. The feedforward component is generated from centrally connected pattern generators (CPGs)in conjunction with nonlinear robot dynamics. A nonlinear integral sliding mode (ISMC) feedback method regulates these rhythmic patterns to adjust robot stability and robustness in the presence of parameter uncertainties and external disturbances. The proposed control strategy is developed, implemented and tested for the spherical robot on both smooth and irregular surfaces. The robot performance is quantified by measuring the stability in roll angle and wheel velocities. Experimental results show that proposed novel strategy is efficient in producing a stable rolling gait and robust control of a spherical robot on different types of surface conditions.

This article discusses the implementation of an infinite-horizon model predictive control approach that is based on representing input and state trajectories by a linear combination of basis functions. An iterative constraint sampling strategy is presented for guaranteeing constraint satisfaction over all times. It will be shown that the proposed method converges. In addition, we will discuss the implementation of the resulting (online) model predictive control algorithm on an unmanned aerial vehicle and provide experimental results. The computational efficiency of the algorithm is highlighted by the fact that a sampling rate of 100 Hz was achieved on an embedded platform.

Tracking for planar objects is an important issue to vision-based robotic applications. In direct visual tracking (DVT) methods, the similarity between two images is often measured through the sum of squared differences (SSD) especially with the efficient second-order minimization (ESM) due to its simplicity and efficiency. However, SSD-based ESM is not robust to illumination changes since it is usually built upon the brightness constancy assumption. Contrast to image brightness, gradient orientations (GO) are invariant to both linear and non-linear illumination changes as verified in practice. Based on GO, we propose an illumination insensitive ESM method for planar object tracking in this paper. In order to introduce GO into the ESM, we generalized the original ESM formulas for multi-dimensional features. In addition, a denoising method based on the Perona-Malik function and a mask image were suggested to improve GO's robustness against image noise and low texture. Our experimental results on dataset for planar objects with illumination changes and a benchmark dataset confirm the proposed method is robust to illumination variations and capable to deal with the general tracking challenges.

Direct visual odometry and Simultaneous Localization and Mapping (SLAM) methods determine camera poses by means of direct image alignment. This optimizes a photometric cost term based on the Lucas-Kanade method. Many recent works use the brightness constancy assumption in the alignment cost formulation and therefore cannot cope with significant illumination changes. Such changes are especially likely to occur for loop closures in SLAM. Alternatives exist which attempt to match images more robustly. In our paper, we perform a systematic evaluation of real-time capable methods. We determine their accuracy and robustness in the context of odometry and of loop closures, both on real images as well as synthetic datasets with simulated lighting changes. We find that for real images, a Census-based method outperforms the others. We make our new datasets available online.

We investigate a variation of the art gallery problem in which a team of mobile guards tries to track an unpredictable intruder in a simply-connected polygonal environment. The guards are deployed based on the strategy initially proposed in [1] which restricts them to move along a specific diagonal within the environment. However, the intruder can move freely within the environment. We define critical regions to generate event-triggered strategies for the guards. Additionally, a hybrid automaton is designed to model the problem, and sufficient conditions are presented for [n/4] guards for persistent surveillance.

We consider the problem of planning whole-body motions for humanoids that must execute loco-manipulation tasks, i.e., manipulation tasks that implicitly require a locomotion phase. The proposed planner builds a tree in configuration-time space by concatenating feasible, collision-free whole-body motions that realize a succession of CoM movement primitives and, at the same time, the assigned manipulation task. To obtain fluid, natural motions we identify three zones of operation, i.e, locomotion, loco-manipulation and manipulation, and we carefully design a mechanism that allows to synchronize the two tasks. The proposed method has been implemented in V-REP for the NAO humanoid and successfully tested in various scenarios of increasing complexity.

In this paper we focus on a mobile platform which physically interacts with a human operator. We detect the contact gestures of a human operator in real-time using a labmade time-of-flight 3D scanner mounted on the platform as well as rotary torque sensors mounted along the drivetrain of its omni-directional wheels. Through the fusion of these two different sensors, touch gestures of an operator are processed inferring information about the body parts in contact and the applied forces. Behaviors that respond to touch-based gestures are programmed a priori, and with the previous sensor data we classify them into a set of known contact gestures that allow the platform to quickly react. We investigate these physical human-robot cooperative functions in a testbed consisting of a sensorized mobile platform and a human operator.

Various assistive machines have been developed to prevent falling accidents of the elderly. In order to achieve advanced support using robot technology, it is important to acquire data or real-time state estimation of user's various motions. However, a lot of expensive and sophisticated sensors utilized to estimate user's state accurately are difficult to use in general households or institutions. In this article, we propose a method to estimate the user's state utilizing a few inexpensive and simple sensors. We focused on CoG (Center of Gravity) to estimate user's state, but when utilizing less sensors than required to calculate the human link model parameters, the position of CoG is underspecified. Then we considered the range of value of unknown parameters to calculate candidates of CoG. The range of CoG candidates can become narrow enough to estimate human state in real-time by properly selecting and placing the sensors. Therefore, the evaluation of CoG candidates allows us to determine where and which sensors to set when designing assistive robots. We firstly selected some sensors which can be generally found on assistive machines, and we created sets of measurements using the number of unknown parameters. From the result of the experiment using a motion capture system, we confirmed that the range of the candidates was considerably narrow when using some of the created measurement sets. We validated the proposed method to estimate user's CoG candidates by actually placing the sensors according to the designed measurement sets and confirmed that the CoG candidates corresponded to those obtained using the motion capture system.

Tool-held hitting tasks, like hammering a nail or striking a ball with a bat, require humans, and robots, to purposely collide and transfer momentum from their limbs to the environment. Due to the vibrational dynamics, every tool has a location where a hit is most efficient results in minimal tool vibrations, and consequently maximum energy transfer to the environment. In sports, this location is often referred to as the “sweet spot” of a bat, or racquet. Our recent neuroscience study suggests that humans optimize hits by using the jerk and torque felt at their hand. Motivated by this result, in this work we first analyze the vibrational dynamics of an end-effector-held bat to understand the signature projected by a sweet spot on the jerk and torque sensed at the end-effector. We then use this analysis to develop a controller for a robotic “baseball hitter”. The controller enables the robot-hitter to iteratively adjust its swing trajectory to ensure that the contact with the ball occurs at the sweet spot of the bat. We tested the controller on the DLR LWR III manipulator with three different bats. Like a human, our robot hitter is able to optimize the energy transfer, specifically maximize the ball velocity, during hits, by using its end effector position and torque sensors, and without any prior knowledge of the shape, size or material of the held bat.

Legged robots can explore unstructured environments more effectively than wheeled robots, but high turning rate tracking is still a challenging problem, particularly on varying surfaces. Previous steering methods with small robots have shown high turn rates, but usually only on a limited set of surfaces. This paper proposes a new method for steering a miniaturized legged robot by cooperation between two robots connected by a compliant joint, creating a 73 gram, 12 legged robot. Detailed design issues and an empirical verification are presented for several cooperation strategies, including changing velocities of the 4 sets of leg triples. The robots use their combined traction forces to turn at better than 50 degrees/sec at 1 m/sec on various surfaces. Closed-loop steering using a differential drive strategy is implemented on the connected robots to track a “figure 8” trajectory on a tile surface.

As robots and other automated systems are introduced to unknown and dynamic environments, robust and adaptive control strategies are required to cope with disturbances, unmodeled dynamics and parametric uncertainties. In this paper, we propose and provide theoretical proofs of a combined L1 adaptive feedback and iterative learning control (ILC) framework to improve trajectory tracking of a system subject to unknown and changing disturbances. The L1 adaptive controller forces the system to behave in a repeatable, predefined way, even in the presence of unknown and changing disturbances; however, this does not imply that perfect trajectory tracking is achieved. ILC improves the tracking performance based on experience from previous executions. The performance of ILC is limited by the robustness and repeatability of the underlying system, which, in this approach, is handled by the L1 adaptive controller. In particular, we are able to generalize learned trajectories across different system configurations because the L1 adaptive controller handles the underlying changes in the system. We demonstrate the improved trajectory tracking performance and generalization capabilities of the combined method compared to pure ILC in experiments with a quadrotor subject to unknown, dynamic disturbances. This is the first work to show L1 adaptive control combined with ILC in experiment.

This study presents the high-precision microinjection of fluorescent micro-gel beads into Caenorhabditis elegans trapped in a suction microchannel. In our previous works, we demonstrated survival microinjection by a conventional micromanipulation technique. However, the focal planes differed between the tip of the microinjection tool and the target axon inside the C. elegans body. To resolve this problem, we here propose a suction microchannel that traps C. elegans during the microinjection operation. The focal plane of the target nerve axon matches that of the fluorescent microbead in the microinjection tool, enabling high-precision microinjection into the interior of the C. elegans body under a microscopic view. In an experimental evaluation, the positioning accuracy of the injection into C. elegans was within the target accuracy (15 μm). The head-flrst navigation alignment of C. elegans along the microchannel was controlled by electrotaxis. Injection of the fluorescent micro-gel bead into the C. elegans body was quantitatively confirmed by confocal microscopy.

Grasping an object in the presence of unpredictable disturbances and with no previous knowledge of the geometry and mass distribution of it represents a significant challenge in Robotics. Also, grasping an object with sufficient force to prevent slippage, whilst not damaging or deforming the shape of the object proves to be still an intricate challenge despite the existence of a huge body of literature on robotic grasping. This paper addresses this challenge by evaluating the performance of a Super Twisting Sliding Mode Control (STSMC) designed for preventing slippage meanwhile causing minimum deformation to the grasped object. The performance of the STSMC is evaluated against the one of a First Order Sliding Mode Controller (FOSMC). Both controllers use grip force and slip feedback to counteract the slippage. Force and slip detection are provided by low-cost, off-the-shelf sensors. Experimental results presented in this paper, show that, when grasping objects, both controllers can robustly overcome external nonlinear disturbances, along with inaccuracies in the system model, preventing slippage and minimising deformation of objects. However, from these results, we observe that the STSMC not only, as expected, eliminates the major drawback of the FOSMC, which is chattering, but also provides more consistent and robust response in terms of slippage and gripping force.

This paper presents an off-board trajectory controller for a range of stride frequencies (2-45 Hz) that enables zero-radius turns and holonomic control on one of the smallest and fastest legged robots, the Harvard Ambulatory MicroRobot (HAMR). An experimental model is used as the basis for control to capture the highly nonlinear response of the robot to input signals. Closed-loop trajectories are performed with an RMS position error at or below 0.3 body lengths (BL) using gaits at speeds up to 6.5 BL/s (29.4 cm/s) for straight-line and sinusoidal trajectories.