ICRA 2021 Papers

In this paper, we present a new algorithm that extends RRT* and RT-RRT* for online path planning in complex, dynamic environments. Sampling-based approaches often perform poorly in environments with narrow passages, a feature common to many indoor applications of mobile robots as well as computer games. Our method extends RRT-based sampling methods to enable the use of an assisting distance metric to improve performance in environments with obstacles. This assisting metric, which can be any metric that has better properties than the Euclidean metric when line of sight is blocked, is used in combination with the standard Euclidean metric in such a way that the algorithm can reap benefits from the assisting metric while maintaining the desirable properties of previous RRT variants - namely probabilistic completeness in tree coverage and asymptotic optimality in path length. We also introduce a new method of targeted rewiring, aimed at shortening search times and path lengths in tasks where the goal shifts repeatedly. We demonstrate that our method offers considerable improvements over existing multi-query planners such as RT-RRT* when using diffusion distance as an assisting metric: finding near-optimal paths with a decrease in search time of several orders of magnitude. Experimental results demonstrate our method offers a reduction of 99.5% in planning times and 9.8% in path lengths over RT-RRT* in a variety of environments.

Model-predictive control (MPC) is an increasingly popular method for controlling complex robotic systems in which optimal control problems are solved on board the robot at real-time rates. However, successful application of MPC depends critically on the performance of the algorithms used to solve the underlying optimization problems. An ideal solver should both leverage the structure of the MPC problem and support efficient "warm starting" so that information from previous solutions can be recycled to speed convergence. We present ALTRO-C, a high-performance solver with both of these properties that utilizes an augmented Lagrangian method to handle general convex conic constraints. We demonstrate the new solver’s superior performance against several existing state-of-the-art solvers on a variety of benchmark control problems formulated as both quadratic and second-order cone programs.

Prior correlation filter (CF)-based tracking methods for unmanned aerial vehicles (UAVs) have virtually focused on tracking in the daytime. However, when the night falls, the trackers will encounter more harsh scenes, which can easily lead to tracking failure. In this regard, this work proposes a novel tracker with anti-dark function (ADTrack). The proposed method integrates an efficient and effective low-light image enhancer into a CF-based tracker. Besides, a target-aware mask is simultaneously generated by virtue of image illumination variation. The target-aware mask can be applied to jointly train a target-focused filter that assists the context filter for robust tracking. Specifically, ADTrack adopts dual regression, where the context filter and the target-focused filter restrict each other for dual filter learning. Exhaustive experiments are conducted on typical dark sceneries benchmark, consisting of 37 typical night sequences from authoritative benchmarks, i.e., UAVDark, and our newly constructed benchmark UAVDark70. The results have shown that ADTrack favorably outperforms other state-of-the-art trackers and achieves a real-time speed of 34 frames/s on a single CPU, greatly extending robust UAV tracking to night scenes.

We introduce ACRONYM, a dataset for robot grasp planning based on physics simulation. The dataset contains 17.7M parallel-jaw grasps, spanning 8872 objects from 262 different categories, each labeled with the grasp result obtained from a physics simulator. We show the value of this large and diverse dataset by using it to train two state-of-the-art learning-based grasp planning algorithms. Grasp performance improves significantly when compared to the original smaller dataset. Data and tools can be accessed at https://sites.google.com/nvidia.com/graspdataset.

Microbubble-induced acoustic microstreaming for efficient on-chip micromanipulation is widely developed in biological applications. However, it is still challenging to simultaneously transport, trap, and rotate single cells using one device in a biocompatible manner, while expensive and bulky traditional acoustic driving system also increases its limitation. This paper presents a portable acoustofluidic device for multifunctional cell manipulation and 3D reconstruction, using acoustically oscillating bottom bubble array. Based on the Arduino-based driving system, multiple bubble-induced microvortices were generated and utilized to achieve multifunctional manipulation in a noninvasive manner. Self-propelled transportation of single or multiple cells is first accomplished by bottom bubble array; Controllable trapping, 3D rotation (in the x-y or x-z plane) of DU145 cells are further performed by every single microbubble. Through experiments, rotation direction, speed and axis can be modulated by tuning the driving frequency and voltage. Finally, 3D cell reconstruction combining imaging processing algorithm with out-of-plane rotation enables a sufficient illustration of cell structures and surface morphology, providing an efficient properties measurement function. All these aspects of this device show great potentials in bioengineering, biophysics and biomedicine.

Modeling the Coulomb Friction Cone in trajectory optimization is typically done by linearizing it along a series of vectors. Often, these vectors define the edges of polyhedral estimations of the cone. This article provides an alternate approach that samples the cone along a vector that satisfies the Maximum Dissipation Principle, which is shown to be significantly more computationally tractable. The proposed technique uses the polar representation of the relative velocity of a contact point on a surface to determine the direction of the resultant friction force and linearizes the friction cone along this vector. This study describes the development of the proposed model. Thereafter, a trajectory optimization experiment was conducted to compare the traditional four-sided polyhedral estimation of the friction cone with the novel method. Compute time and computational complexity were used as performance metrics in this study. Results from these experiments indicate that the proposed method reduced the compute time by 39.13% and a 57.00% reduction in inequality constraints when compared to the 4-sided polyhedral estimation.

Flapping-wing insect-scale robots (<500 mg) offer enormous potential advantages over larger robots in applications such as agricultural support, environmental monitoring, and exploration of hazardous locations or extra-terrestrial space. Scaling laws that confer advantages to smaller robots, such as reduced materials cost and lower power requirements, also pose miniaturization challenges. Here we address the challenge of supplying a high voltage, oscillating signal to piezoelectric actuators that can be modulated to vary wing thrust. We present a system capable of modulating thrust for the generation of forces and torques required for control of flapping-wing insect sized robots. The power electronics unit with boost converter and two drivers is 90 mg including optional 15 mg output storage capacitor (not including 8 mg MCU). The boost converter operates at 30–40% efficiency at 240 V output under driver load and supplied with 7 V, and the driver produces sinusoidal wing flapping signals at typical operating output voltages of 160–220 Vpp with <14% total harmonic distortion. The system can linearly modulate measured thrust over 70% of the tested amplitude range from 40–200 Vpp. The thrust modulation reported here is necessary to realize controlled flight using on-board power systems, instead of externally provided signals via a wire tether.

Camera motion estimation, such as relative pose estimation and absolute pose estimation, are fundamental problems in computer vision and robotics. To obtain the motion parameters, classical methods rely on studying the properties of the geometric matrices, e.g., rotation matrix, essential matrix, homography matrix. The well known five-point algorithm was successfully derived using the singular constraint and trace constraints on the essential matrix. However, finding all the algebraic constraints is not always trivial for some recent problems. In this paper, we propose a simple and general technique to find complete algebraic constraints so that we can derive efficient algorithms. We show that using the quaternion to formulate the rotation matrix we can eliminate any unknowns from the original equations and obtain constraints on the rest of the unknowns based on Gröbner basis. We demonstrate that this approach can be applied to almost all the camera motion estimation and show its improvement compared to the existing methods. Further more, based on this elimination technique, we exploit new constraints for the relative pose estimation with gravity prior, and derive a new globally optimal algorithm to this problem. We compare our algorithm with the state-of-the-art methods on both synthetic and real-world data, and show the benefits including accuracy and efficiency.

3D pointing devices are indispensable in virtual reality (hereafter VR) and human-robot interaction scenarios. Existing devices are cumbersome or non-immersive or have a limited volume of operation. Hand gesture-based interfaces do not suffer from these problems and can be used for 3D pointing purposes. However, there is a lack of robust, accurate hand gesture-based pointing techniques which can be attributed to the non-existence of large and accurate data-set for the same. To overcome this barrier, we propose a data-set consisting of depth images with a large number (107000) of samples collected from 11 subjects, with accurate ground-truth and adequate variation in the orientation and distance of the hand w.r.t. the camera. We propose a 3D convolutional neural network based technique that works on the proposed data-set and achieves an accuracy of 94.49% for an angle error threshold of 10 degrees. The proposed data-set may be used for developing more accurate, robust, less computationally expensive methods.

This paper presents a wheeled V-shaped in-pipe robot in which the two outputs of the wheel shaft and roll joint are driven solely by a single actuator input. This underactuation is generated by a simple miter gear mechanism. Generally, to control two movements easily, one of the outputs of the underactuated mechanism is constrained by the resilience force of springs or by the friction force. However, this complicates the control of each output. In this study, a one-way clutch is installed to completely constrain one of the outputs (wheel movement). By using this clutch, the proposed mechanism enables a hemispherical wheel to switch between pitch and roll rotations by selecting the drive direction of a single motor. The one-way clutch constrains the wheels to rotate in only one direction. To take advantage of this constraint, the robot changes its direction of movement between forward and backward by using the rolling movement of the robot. After describing the configuration of the proposed robot and a roll-angle model of the robot, experiments are conducted in straight pipes, bending pipes, and an out-of-plane double elbow.

Given a two-dimensional polygonal space, the multi-robot visibility-based pursuit-evasion problem tasks several pursuer robots with the goal of establishing visibility with an arbitrarily fast evader. The best known complete algorithm for this problem takes time doubly exponential in the number of robots. However, sampling-based techniques have shown promise in generating feasible solutions in these scenarios. One of the primary drawbacks to employing existing sampling-based methods is that existing algorithms have long execution times and high failure rates for complex environments. This paper addresses that limitation by proposing a new algorithm that takes an environment as its input and returns a joint motion strategy which ensures that the evader is captured by one of the pursuers. Starting with a single pursuer, we sequentially construct Sample-Generated Pursuit-Evasion Graphs to create such a joint motion strategy. This sequential graph structure ensures that our algorithm will always terminate with a solution, regardless of the complexity of the environment. We describe an implementation of this algorithm and present quantitative results that show significant improvement in comparison to the existing algorithm.

This paper describes a versatile platform for swarm robotics research. It integrates multiple pheromone communication with a dynamic visual scene along with real time data transmission and localization of multiple-robots. The platform has been built for inquiries into social insect behavior and bio-robotics. By introducing a new research scheme to coordinate olfactory and visual cues, it not only complements current swarm robotics platforms which focus only on pheromone communications by adding visual interaction, but also may fill an important gap in closing the loop from bio-robotics to neuroscience. We have built a controllable dynamic visual environment based on our previously developed ColCOSΦ (a multi-pheromones platform) by enclosing the arena with LED panels and interacting with the micro mobile robots with a visual sensor. In addition, a wireless communication system has been developed to allow transmission of real-time bi-directional data between multiple micro robot agents and a PC host. A case study combining concepts from the internet of vehicles (IoV) and insect-vision inspired model has been undertaken to verify the applicability of the presented platform, and to investigate how complex scenarios can be facilitated by making use of this platform.

Probabilistic regression techniques in control and robotics applications have to fulfill different criteria of data-driven adaptability, computational efficiency, scalability to high dimensions, and the capacity to deal with different modalities in the data. Classical regressors usually fulfill only a subset of these properties. In this work, we extend seminal work on Bayesian nonparametric mixtures and derive an efficient variational Bayes inference technique for infinite mixtures of probabilistic local polynomial models with well-calibrated certainty quantification. We highlight the model’s power in combining data-driven complexity adaptation, fast prediction, and the ability to deal with discontinuous functions and heteroscedastic noise. We benchmark this technique on a range of large real-world inverse dynamics datasets, showing that the infinite mixture formulation is competitive with classical Local Learning methods and regularizes model complexity by adapting the number of components based on data and without relying on heuristics. Moreover, to showcase the practicality of the approach, we use the learned models for online inverse dynamics control of a Barrett-WAM manipulator, significantly improving the trajectory tracking performance.

This paper presents a new variable stiffness actuator based on a second-order lever mechanism which has wide stiffness regulation range. By employing a novel symmetric structure design and improving the load capacity of the stiffness regulation module, the proposed actuator also shows well performance in load capacity, stiffness regulation response, and elastic hysteresis. On this basis, a variable stiffness actuated manipulator is developed. The experimental results demonstrate that the presented manipulator possesses abilities in fast stiffness tracking, shock-absorbing and explosive movement. It is also verified that the manipulator can withstand accidental impact, which illustrates the structure stability of the proposed design.

Interactive (mechanical) impedance and finger fatigues are important topics, which have not been well investigated. To tackle this problem, we developed a soft lightweight (0.25 kg) finger exoskeleton (TIE-EXO) for quantifying interactive impedance and finger fatigue. A resist-as-needed (RAN) controller was used to produce variable resistance in fingers’ exercises. The TIE-EXO’s feedback and RAN’s parameters were applied to quantify the relationship between interactive impedance and finger fatigue. This quantification was validated in the index and middle fingers of three subjects. This validation shows that the RAN control enables the TIE-EXO to produce online resistance adaptations to different subjects and finger fatigue. Moreover, it indicates a variation and invariance in finger impedance control. We argue that the proposed method provides a novel way for investigating interactive impedance and finger fatigue.

During the co-presence of human workers and robots, measures are required to avoid injuries from undesired contacts. Capacitive Proximity Sensors (CPSs) offer a cost-effective solution to cover the entire robot manipulator with fast close-range perception for HRC tasks, closing the perception gap between tactile detection and mid-range perception. CPSs do not suffer from occlusion and compared to pure tactile or force sensing, they react earlier and allow increasing the operating speed of Collaborative Robots (Cobots) while still maintaining safety. However, since capacitive coupling to obstacles varies with their distance, shape and material properties, the projection from capacitance to actual distances is a general problem. In this work, we propose an universal benchmark test procedure for fellow researchers to evaluate their CPSs. Considering ISO/TS 15066 for Power and Force Limiting (PFL) as a reference, we derive the requirements for the specified body regions and propose a method for determining the operation speed to comply with PFL based on a pre-defined detection threshold. Finally, the benchmark test procedure is evaluated on three different concepts of CPSs from the contributed researchers, demonstrating the general applicability.

This work presents a framework for the simultaneous optimization of motors, transmissions, and mechanisms of different joints of robotic legs with the goal of achieving an energy efficient, precisely controllable and stable locomotion in dynamic environments. This unified framework allowed us to introduce and formulate new performance metrics for the separate evaluation of the system's stabilizing ability during stance and swing. Moreover, through a case study, this design optimization framework was applied to an anthropomorphic robot leg model and the optimal actuation configurations for the leg were obtained. This case study also helped us investigate the relationships among our three objectives (energy efficiency, and stance and swing control). It was shown that while in some cases a clear trade-off exists, it is not always valid and as such, careful consideration of all three objectives is necessary.

We introduce an origami-reinforced parallel continuum robot which is capable of maintaining the orientation of the end effector regardless of the bending shape. The cross-routing tendons provide an effective actuation of the robot because the constant length of the backbones prevent the actuation of parallel arrangement of tendons. We utilise the arclength relationship of parallel curves to show that parallel backbones of equal lengths and constant spacing between the backbones enable the constant orientation of the end effector. The origami shell is introduced to increase the torsional stiffness of the continuum robot and minimise any twisting. This leads to planar and parallel bending of all backbones. Planar quintic Pythagorean Hodograph curves are utilised for shape reconstruction, which is more accurate as compared to the curves with piecewise constant curvatures. The concept of this continuum robot and the accuracy of the reconstruction are validated experimentally.

Unmanned aerial vehicles (UAVs) are finding their way into offshore applications. In this work, we postulate an original system that entails a marine locomotive quadrotor UAV that manipulates the velocity of a floating buoy by means of a cable. By leveraging the advantages of UAVs relative to high speed, maneuverability, ease of deployment, and wide field of vision, the proposed UAV−buoy system paves the way in front of a variety of novel applications. The dynamic model that couples the buoy, UAV, cable, and water environment is presented using the Euler-Lagrange method. A stable control system design is proposed to manipulate the forward-surge speed of the buoy under two constraints: maintaining the cable in a taut state, and keeping the buoy in contact with the water surface. Polar coordinates are used in the controller design process to attain correlated effects on the tracking performance, whereby each control channel independently affects one control parameter. This results in improved performance over traditional Cartesian-based velocity controllers, as demonstrated via numerical simulations in wave-free and wavy seas.