ICRA 2021 论文列表

Nowadays, various algorithms based on the Rapidly-exploring Random Tree (RRT) methods are utilized to solve motion planning problems. Based on the RRT*, we developed a novel reconnection method that enables the planner to directly generate a smooth curved trajectory. Meanwhile, kinodynamic constraints of the robots are considered to generate the control input, which improves the feasibility of the algorithm. The trajectory planned by the Smooth-RRT* is significantly suitable for the non-holonomic robots. Planning tests are conducted in four scenarios to demonstrate performance of the proposed algorithm in comparison with the original RRT* and kinodynamic-RRT (Kino-RRT). Smooth-RRT* yields shorter and smoother planned path in all the scenarios compared with the Kino-RRT. It finds a solution with fewer expansion nodes than the RRT* under the same time consumption. The results demonstrate that the proposed algorithm can generate a smooth trajectory satisfied with the kinodynamic constraints and ensure the asymptotic optimality.

Continuum arms, with their mix of compliance, payload, safety, and manipulability, are perfectly suited to serve as co-robots, and their applications range from industry and manufacturing to human healthcare. Their hyper-redundancy serves as their most significant challenge for path planning and path planning approaches commonly used with rigid-link robots, such as inverse kinematics, that fail to provide reliable trajectories for continuum arms. We propose an Inverse Kinematics-based approach to address the limitations of previously-proposed Kinematics-based approaches. Using this new approach, we are able to efficiently generate very rich sets of configurations, which, in turn, lead to smooth path planning for such continuum manipulators. To validate the smoothness of the paths generated by our approach, we apply dynamics constraints to the generated trajectories. We show that, when tracked by a controller, the paths that are generated using the proposed approach are much smoother than previously-proposed Kinematics-based approaches: The proposed approach allows the continuum arm to traverse the trajectories very accurately and in time less than half of that taken by previous (reliable) path planning approaches.

Ability to generate intelligent and generalizable facial expressions is essential for building human-like social robots. At present, progress in this field is hindered by the fact that each facial expression needs to be programmed by humans. In order to adapt robot behavior in real time to different situations that arise when interacting with human subjects, robots need to be able to train themselves without requiring human labels, as well as make fast action decisions and generalize the acquired knowledge to diverse and new contexts. We addressed this challenge by designing a physical animatronic robotic face with soft skin and by developing a vision-based self-supervised learning framework for facial mimicry. Our algorithm does not require any knowledge of the robot's kinematic model, camera calibration or predefined expression set. By decomposing the learning process into a generative model and an inverse model, our framework can be trained using a single motor babbling dataset. Comprehensive evaluations show that our method enables accurate and diverse face mimicry across diverse human subjects.

Solar-powered actuation of a 15 mN electrostatic MEMS gripper was demonstrated while wirelessly triggered by IEEE 802.15.4 RF signals. The solar-powered gripper was shown to actuate at a rate of 640 um/s. The complete system is composed of three capacitors and three chips: MEMS gripper, microprocessor/crystal-free radio, and solar cell array/high voltage buffer. Control signals for the electrostatic inchworm motors originate from the 3×2×0.3 mm3 chip with an ARM Cortex-M0 microprocessor and are passed through 119 V high voltage buffers. Power for all components, including the crystal-free radio, microprocessor, and 119 V buffers, is supplied by a multi-output array of solar cells on a CMOS SOI chip under 200 mW/cm2 irradiation.

This work proposes a quaternion-based sliding variable that describes exponentially convergent error dynamics for any forward complete desired attitude trajectory. The proposed sliding variable directly operates on the non-Euclidean space formed by quaternions and explicitly handles the double covering property to enable global attitude tracking when used in feedback. In-depth analysis of the sliding variable is provided and compared to others in the literature. Several feedback controllers including nonlinear PD, robust, and adaptive sliding control are then derived. Simulation results of a rigid body with uncertain dynamics demonstrate the effectiveness and superiority of the approach.

It is inevitable for human to bear the gravitational and inertial force when carrying loads. The impact force exerted on human body is originated from the inertial force which can increase the energy expenditure and cause injury to human body. This paper proposes a semi-active hover backpack with controllable air damper to minimize the inertial force. The skyhook damper model of hover backpack is established which is the dynamic target of the practical backpack. Sliding mode control is designed to eliminate the tracking error and the effectiveness of the control method is analyzed. Simulation and experiment are conducted and comparative results are stated. The results demonstrate that the semi-active hover backpack with sliding mode control can reduce the inertial force.

Shared autonomy enables robots to infer user intent and assist in accomplishing it. But when the user wants to do a new task that the robot does not know about, shared autonomy will hinder their performance by attempting to assist them with something that is not their intent. Our key idea is that the robot can detect when its repertoire of intents is insufficient to explain the user's input, and give them back control. This then enables the robot to observe unhindered task execution, learn the new intent behind it, and add it to this repertoire. We demonstrate with both a case study and a user study that our proposed method maintains good performance when the human's intent is in the robot's repertoire, outperforms prior shared autonomy approaches when it isn't, and successfully learns new skills, enabling efficient lifelong learning for confidence-based shared autonomy.

Haptic guidance can improve accuracy and dexterity during the teleoperation of a robot, but only if the model of the task used to provide the assistance is accurate. In medical robotics, the registration of a task from pre-operative planning from medical images to the robot’s task-space can be erroneous. Additionally, the deformability of the environment can require online correction of a planned task. Therefore, we propose a method to update the geometry and the registration of a pathfollowing task online. This model is simultaneously used to physically guide the user during the teleoperation. Experimental results obtained on a haptic interface show the validity of the approach for a simulated 2D task.

Simultaneous assembly of multiple objects is a key technology to form solid connections among objects to get compact structures in precision assembly and micro-assembly. Dramatically different from traditional assembly of two objects, the interaction among multiple objects is more complicated on analysis and control. During simultaneous assembly of multiple objects, there are multiple mutually effected contact surfaces, and multiple force sensors are needed to perceive the interaction status. In this paper, a coordinated micro-robot manipulation strategy is proposed for simultaneous assembly problem, which is based on microscopic vision and force information. Taking simultaneous assembly of three objects as an instance, the proposed method is well articulated, including calibration of assembly system, force analysis for each contacting surface, and insertion control strategy for assembly process. The proposed method is applicable also to case with more objects. Experiment results demonstrate effectiveness of the proposed method.

This paper presents a semi-supervised framework for multi-level description learning aiming for robust and accurate camera relocalization across large perception variations. Our proposed network, namely DLSSNet, simultaneously learns weakly-supervised semantic segmentation and local feature description in the hierarchy. Therefore, the augmented descriptors, trained in an end-to-end manner, provide a more stable high-level representation for local feature dis-ambiguity. To facilitate end-to-end semantic description learning, the descriptor segmentation module is proposed to jointly learn semantic descriptors and cluster centers using standard semantic segmentation loss. We show that our model can be easily fine-tuned for domain-specific usage without any further semantic annotations, instead, requiring only 2D-2D pixel correspondences. The learned descriptors, trained with our proposed pipeline, can boost the cross-season localization performance against other state-of-the-arts.

This paper develops a probabilistic simultaneous estimation and modeling (SEAM) framework for estimating a robot’s state and correcting its motion model parameters. This is done by incorporating model uncertainty in state prediction and correcting parameters via optimization. In the proposed technique, belief about a state being estimated is represented by arbitrary multi-dimensional non-Gaussian probability distribution functions. The approach is validated in proof-of-concept for second-order simulated systems whose models are poorly estimated. Given sufficient state observations, the proposed framework reliably reduces and usually converges model parameter error. In comparison with existing advanced estimators, robotic state estimation is enhanced under this framework when model uncertainty is high and state belief is highly unstructured and non-Gaussian. This work holds promise for challenging robotic localization, estimation, and prediction problems across many complex domains.

Tactile sensing has seen a rapid adoption with the advent of vision-based tactile sensors. Vision-based tactile sensors provide high resolution, compact and inexpensive data to perform precise in-hand manipulation and human-robot interaction. However, the simulation of tactile sensors is still a challenge. In this paper, we built the first fully general optical tactile simulation system for a GelSight sensor using physics based rendering techniques. We propose physically accurate light models and show in-depth analysis of individual components of our simulation pipeline. Our system outperforms previous simulation techniques qualitatively and quantitative on image similarity metrics. Our code and experimental data is open-sourced at \color{Fuchsia} {\text{project page}}\color{Fuchsia} {\text{project page}} project page.

Robust and reliable ego-motion is a key component of most autonomous mobile systems. Many odometry estimation methods have been developed using different sensors such as cameras or LiDARs. In this work, we present a resilient approach that exploits the redundancy of multiple odometry algorithms using a 3D LiDAR scanner and a monocular camera to provide reliable state estimation for autonomous vehicles. Our system utilizes a stack of odometry algorithms that run in parallel. It chooses from them the most promising pose estimation considering sanity checks using dynamic and kinematic constraints of the vehicle as well as a score computed between the current LiDAR scan and a locally built point cloud map. In this way, our method can exploit the advantages of different existing ego-motion estimating approaches. We evaluate our method on the KITTI Odometry dataset. The experimental results suggest that our approach is resilient to failure cases and achieves an overall better performance than individual odometry methods employed by our system.

In this work we present a simple end-to-end trainable machine learning system capable of realistically simulating driving experiences. This can be used for verification of self-driving system performance without relying on expensive and time-consuming road testing. In particular, we frame the simulation problem as a Markov Process, leveraging deep neural networks to model both state distribution and transition function. These are trainable directly from the existing raw observations without the need of any handcrafting in the form of plant or kinematic models. All that is needed is a dataset of historical traffic episodes. Our formulation allows the system to construct never seen scenes that unfold realistically reacting to the self-driving car’s behaviour. We train our system directly from 1,000 hours of driving logs and measure both realism, reactivity of the simulation as the two key properties of the simulation. At the same time we apply the method to evaluate performance of a recently proposed state-of-the-art ML planning system [1] trained from human driving logs. We discover this planning system is prone to previously unreported causal confusion issues that are difficult to test by non-reactive simulation. To the best of our knowledge, this is the first work that directly merges highly realistic data-driven simulations with a closed loop evaluation for self-driving vehicles. We make the data, code, and pre-trained models publicly available to further stimulate simulation development.

As learning-based approaches progress towards automating robot controllers design, transferring learned policies to new domains with different dynamics (e.g. sim-to-real transfer) still demands manual effort. This paper introduces SimGAN, a framework to tackle domain adaptation by identifying a hybrid physics simulator to match the simulated trajectories to the ones from the target domain, using a learned discriminative loss to address the limitations associated with manual loss design. Our hybrid simulator combines neural networks and traditional physics simulation to balance expressiveness and generalizability, and alleviates the need for a carefully selected parameter set in System ID. Once the hybrid simulator is identified via adversarial reinforcement learning, it can be used to refine policies for the target domain, without the need to interleave data collection and policy refinement. We show that our approach outperforms multiple strong baselines on six robotic locomotion tasks for domain adaptation.

Tactile sensing is critical for robotic grasping and manipulation of objects under visual occlusion. However, in contrast to simulations of robot arms and cameras, current simulations of tactile sensors have limited accuracy, speed, and utility. In this work, we develop an efficient 3D finite element method (FEM) model of the SynTouch BioTac sensor using an open-access, GPU-based robotics simulator. Our simulations closely reproduce results from an experimentally-validated model in an industry-standard, CPU-based simulator, but at 75x the speed. We then learn latent representations for simulated BioTac deformations and real-world electrical output through self-supervision, as well as projections between the latent spaces using a small supervised dataset. Using these learned latent projections, we accurately synthesize real-world BioTac electrical output and estimate contact patches, both for unseen contact interactions. This work contributes an efficient, freely-accessible FEM model of the BioTac and comprises one of the first efforts to combine self-supervision, cross-modal transfer, and sim-to-real transfer for tactile sensors.

In this study, we present a method to grasp diverse unseen real-world objects using an off-policy actor-critic deep reinforcement learning (RL) with the help of a simulation and the use of as little real-world data as possible. Actor-critic deep RL is unstable and difficult to tune when a raw image is given as an input. Therefore, we use state representation learning (SRL) to make actor-critic RL feasible for visual grasping tasks. Meanwhile, to reduce visual reality gap between simulation and reality, we also employ a typical pixel-level domain adaptation that can map simulated images to realistic ones. In our method, as the SRL model is a common preprocessing module for simulated and real-world data, we perform SRL using real and adapted images. This pixel-level domain adaptation enables the robot to learn grasping skills in a real environment using small amounts of real-world data. However, the controller trained in the simulation should adapt to the real world efficiently. Hence, we propose a method combining a typical pixel-level domain adaptation and the proposed SRL model, where we perform SRL based on a feature-level domain adaptation. In evaluations of vision-based robotics grasping tasks, we show that the proposed method achieves a substantial improvement over a method that only employs a pixel-level or domain adaptation.

We study the problem of realizing the full spectrum of bipedal locomotion on a real robot with sim-to-real reinforcement learning (RL). A key challenge of learning legged locomotion is describing different gaits, via reward functions, in a way that is intuitive for the designer and specific enough to reliably learn the gait across different initial random seeds or hyperparameters. A common approach is to use reference motions (e.g. trajectories of joint positions) to guide learning. However, finding high-quality reference motions can be difficult and the trajectories themselves narrowly constrain the space of learned motion. At the other extreme, reference-free reward functions are often underspecified (e.g. move forward) leading to massive variance in policy behavior, or are the product of significant reward-shaping via trial-and-error, making them exclusive to specific gaits. In this work, we propose a reward-specification framework based on composing simple probabilistic periodic costs on basic forces and velocities. We instantiate this framework to define a parametric reward function with intuitive settings for all common bipedal gaits - standing, walking, hopping, running, and skipping. Using this function we demonstrate successful sim-to-real transfer of the learned gaits to the bipedal robot Cassie, as well as a generic policy that can transition between all of the two-beat gaits.

We reformulate the signal temporal logic (STL) synthesis problem as a maximum a-posteriori (MAP) inference problem. To this end, we introduce the notion of random STL (RSTL), which extends deterministic STL with random predicates. This new probabilistic extension naturally leads to a synthesis-as-inference approach. The proposed method allows for differentiable, gradient-based synthesis while extending the class of possible uncertain semantics. We demonstrate that the proposed framework scales well with GPU-acceleration, and present realistic applications of uncertain semantics in robotics that involve target tracking and the use of occupancy grids.