ICRA 2022 论文列表

Robotic airships offer significant advantages in terms of safety, mobility, and extended flight times. However, their highly restrictive weight constraints pose a major challenge regarding the available computational resources to perform the required control tasks. Neuromorphic computing stands for a promising research direction for addressing such problem. By mimicking the biological process for transferring information between neurons using spikes or impulses, spiking neural networks (SNNs) allow for low power consumption and asynchronous event-driven processing. In this paper, we propose an evolved altitude controller based on an SNN for a robotic airship which relies solely on the sensory feedback provided by an airborne radar. Starting from the design of a lightweight, low-cost, open-source airship, we also present an SNN-based controller architecture, an evolutionary framework for training the network in a simulated environment, and a control strategy for ameliorating the gap with reality. The system's performance is evaluated through real-world experiments, demonstrating the advantages of our approach by comparing it with an artificial neural network and a linear controller. The results show an accurate tracking of the altitude command with an efficient control effort.

This article studies the tracking control of a quadrotor unmanned aerial vehicle (UAV) under time-varying external disturbances. An event-triggered sliding mode control (SMC) strategy is proposed by introducing a new triggering condition form of desired trajectory, quadrotor position, and velocity. In the sense of Lyapunov theory, the stability of the entire closed-loop control system is analyzed, and it is proved that the tracking error is adjusted to an adjustable set around zero. We show that the Zeno phenomenon can be avoided; that is, a positive minimum inter-event time is assured. One of the salient features of the proposed strategy is that it can reduce the update frequency of the control efforts, thereby ensuring desirable tracking performance under limited communication bandwidth. Comparative simulation and experimental results are provided to show the efficacy of our framework.

To certify UAV operations in populated areas, risk mitigation strategies - such as Emergency Landing (EL) - must be in place to account for potential failures. EL aims at reducing ground risk by finding safe landing areas using on-board sensors. The first contribution of this paper is to present a new EL approach, in line with safety requirements introduced in recent research. In particular, the proposed EL pipeline includes mechanisms to monitor learning based components during execution. This way, another contribution is to study the behavior of Machine Learning Runtime Monitoring (MLRM) approaches within the context of a real-world critical system. A new evaluation methodology is introduced, and applied to assess the practical safety benefits of three MLRM mechanisms. The proposed approach is compared to a default mitigation strategy (open a parachute when a failure is detected), and appears to be much safer.

Most human activities in daily living or professional work rely on upper body motion. Measuring upper body motion is essential for many applications such as health evaluation, rehabilitation, human power augmentation, skill transferring, etc. Computer vision-based systems have been widely used to directly capture upper limb motion but are usually constrained in a restricted area. Wearable sensors such as inertial measurement units (IMUs) are promising to enable ambulant and out-of-lab measurements but also suffer from issues such as magnetic distortion and drifting. Some visual-inertial systems have been proposed recently to fuse these two complementary measurements but mostly apply in a restricted area. In this paper, we propose a fully wearable egocentric visual-inertial system to estimate the upper-limb pose. Magnetometers are not used to allow the system to work in complex industrial and daily living scenarios or to be integrated with motorized assistive devices. Methods to automatically calibrate the sensor-to-segment alignment and estimate upper body motion is presented and validated with an optical motion capture system. Experimental results showed the system can estimate the joint angles without drift and obtain accurate wrist position even with occlusion, verifying the efficacy of the proposed system and method.

Inertial Navigation Systems (INS) are a key technology for autonomous vehicles applications. Recent advances in estimation and filter design for the INS problem have exploited geometry and symmetry to overcome limitations of the classical Extended Kalman Filter (EKF) approach that formed the mainstay of INS systems since the mid-twentieth century. The industry standard INS filter, the Multiplicative Extended Kalman Filter (MEKF), uses a geometric construction for attitude estimation coupled with classical Euclidean construction for position, velocity and bias estimation. The recent Invariant Extended Kalman Filter (IEKF) provides a geometric framework for the full navigation states, integrating attitude, position and velocity, but still uses the classical Euclidean construction to model the bias states. In this paper, we use the recently proposed Equivariant Filter (EqF) framework to derive a novel observer for biased inertial-based navigation in a fully geometric framework. The introduction of virtual velocity inputs with associated virtual bias leads to a full equivariant symmetry on the augmented system. The resulting filter performance is evaluated with both simulated and real-world data, and demonstrates increased robustness to a wide range of erroneous initial conditions, and improved accuracy when compared with the industry standard Multiplicative EKF (MEKF) approach.

In the presence of fast motion, point clouds obtained from mechanical spinning LiDAR can be easily distorted due to the slow scanning speed of the LiDAR. Existing LiDAR-only odometry algorithms generally ignore this distortion or compensate by linearly interpolating the estimated relative motion between scans. However, when there are abrupt and nonlinear motion changes, the linear interpolation method poorly compensates for the distortions, which can cause significant drift in motion estimates. In this work, we present a LiDAR-only odometry algorithm that estimates motion by slicing LiDAR scans into shorter times to compensate more agilely for point cloud distortions. Observations from only one small scan slice inevitably lack spatial uniqueness, so the multimodal problem needs to be addressed. For LiDAR-only odometry with small scan slices, we introduce the ensemble Kalman filter, a kind of Monte Carlo-based Bayesian filter. The proposed method makes it possible to perform odometry with only a very narrow field of view (FoV), and the robustness to point cloud distortion is improved. We demonstrate the effectiveness of the proposed method through Monte Carlo simulations and several tests with fast-moving scenarios. The experimental results prove the possibility of odometry with a very narrow FoV of down to 10 degrees and robustness against motion distortion.

The gaits of locomoting systems are typically designed to maximize some sort of efficiency, such as cost of transport or speed. Equally important is the ability to modulate such a gait to effect turning maneuvers. For drag-dominated systems, geometric mechanics provides an elegant and practical framework for both ends—gait design and gait modulation. Within this framework, “constraint curvature” maps can be used to approximate the net displacement of robotic systems over cyclic gaits. Gait optimization is made possible under a previously reported “soap-bubble” algorithm. In this work, we propose both local and global gait morphing algorithms to modify a nominal gait to provide single-parameter steering control. Using a simplified swimmer, we numerically compare the two approaches and show that for modest turns, the local approach, while suboptimal, nevertheless proves effective for steering control. A potential advantage of the local approach is that it can be readily applied to soft robots or other systems where local approximations to the constraint curvature can be garnered from data, but for which obtaining an exact global model is infeasible.

This paper presents a physical interface for collaborative mobile manipulators in industrial manufacturing and logistics applications. The proposed work builds on our earlier MOCA-MAN interface, through which an operator could be physically coupled to a mobile manipulator to be assisted in performing daily activities. The previous interface was based on a magnetic clamp attached to one arm of the user for the coupling stage, and a bracelet based on EMG sensors on the other arm for human-robot communication via gestures. The new interface instead presents the following additions: i) An industrial-like design that allows the worker to couple/decouple easily and to operate mobile manipulators locally; ii) A simplistic communication channel via a simple buttons board that allows controlling the robot with one hand only; iii) The interface offers enhanced loco-manipulation capabilities that do not compromise the worker mobility. In addition, an experimental evaluation with six human subjects is carried out to analyze the enhanced locomotion and flexibility of the proposed interface in terms of mobility constraint, usability, and physical load reduction.

Recent Multi-Agent Deep Reinforcement Learning approaches factorize a global action-value to address non-stationarity and favor cooperation. These methods, however, hinder exploration by introducing constraints (e.g., additive value-decomposition) to guarantee the factorization. Our goal is to enhance exploration and improve sample efficiency of multi-robot mapless navigation by incorporating a periodical Evolutionary Policy Search (EPS). In detail, the multi-agent training “specializes” the robots' policies to learn the collision avoidance skills that are mandatory for the task. Concurrently, in this work we propose the use of Evolutionary Algorithms to explore different regions of the policy space in an environment with only a single robot. The idea is that core navigation skills, originated by the multi-robot policies using mutation operators, improve faster in the single-robot EPS. Hence, policy parameters can be injected into the multi-robot setting using crossovers, leading to improved performance and sample efficiency. Experiments in tasks with up to 12 robots confirm the beneficial transfer of navigation skills from the EPS to the multi-robot setting, improving the performance of prior methods.

This paper presents algorithms for performing data-driven reachability analysis under temporal logic side information. In certain scenarios, the data-driven reachable sets of a robot can be prohibitively conservative due to the inherent noise in the robot's historical measurement data. In the same scenarios, we often have side information about the robot's expected motion (e.g., limits on how much a robot can move in a one-time step) that could be useful for further specifying the reachability analysis. In this work, we show that if we can model this side information using a signal temporal logic (STL) fragment, we can constrain the data-driven reachability analysis and safely limit the conservatism of the computed reachable sets. Moreover, we provide formal guarantees that, even after incorporating side information, the computed reachable sets still properly over-approximate the robot's future states. Lastly, we empirically validate the prac-ticality of the over-approximation by computing constrained, data-driven reachable sets for the Small- Vehicles-for-Autonomy (SVEA) hardware platform in two driving scenarios.

Collision-free navigation while moving amongst static and dynamic obstacles with a limited sensor range is still a great challenge for modern mobile robots. Therefore, the ability to avoid collisions with obstacles in crowded, partially observable environments is one of the most important indicators to measure the navigation performance of a mobile robot. In this paper, we propose a novel deep reinforcement learning architecture that combines a spatial graph and attention rea-soning to tackle this problem. We take the relative positions and velocities of observed humans as nodes of the spatial graph and robot-human pairs as nodes of the attention graph to capture the spatial relations between the robot and the humans. In this way, our approach enhances the modeling of the relationship between the moving robot, static obstacles, and the people in the surrounding. As a result, our proposed navigation framework significantly outperforms state-of-the-art approaches [1], [2] in crowded scenarios when the robot has only a limited sensor range in terms of a reduced collision rate. Furthermore, we realize a seriously decreased training time by applying parallel Double Deep Q-Learning.

Despite its importance, unsupervised domain adaptation (UDA) on LiDAR semantic segmentation is a task that has not received much attention from the research community. Only recently, a completion-based 3 DD method has been proposed to tackle the problem and formally set up the adaptive scenarios. However, the proposed pipeline is complex, voxel-based and requires multi-stage inference, which inhibits it for real-time inference. We propose a range image-based, effective and efficient method for solving UDA on LiDAR segmentation. The method exploits class prototypes from the source domain to pseudo label target domain pixels, which is a research direction showing good performance in UDA for natural image semantic segmentation. Applying such approaches to LiDAR scans has not been considered because of the severe domain shift and lack of pre-trained feature extractor that is unavailable in the LiDAR segmentation setup. However, we show that proper strategies, including reconstruction-based pre-training, enhanced prototypes, and selective pseudo labeling based on distance to prototypes, is sufficient enough to enable the use of prototypical approaches. We evaluate the performance of our method on the recently proposed LiDAR segmentation UDA scenarios. Our method achieves remarkable performance among contemporary methods.

Although manipulation capabilities of aerial robots greatly improved in the last decade, only few works addressed the problem of aerial physical interaction with dynamic environments, proposing strongly model-based approaches. However, in real scenarios, modeling the environment with high accuracy is often impossible. In this work, we aim at developing a control framework for Omnidirectional Micro Aerial Vehicles (OMAVs) for reliable physical interaction tasks with articulated and movable objects in the presence of possibly unforeseen disturbances, and without relying on an accurate model of the environment. Inspired by previous applications of energy-based controllers for physical interaction, we propose a passivity-based impedance and wrench tracking controller in combination with a momentum-based wrench estimator. This is combined with an energytank framework to guarantee the stability of the system, while energy and power flow-based adaptation policies are deployed to enable safe interaction with any type of passive environment. The control framework provides formal guarantees of stability, which is validated in practice considering the challenging task of pushing a cart of unknown mass, moving on a surface of unknown friction, as well as subjected to unknown disturbances. For this scenario, we present, evaluate and discuss three different policies.

Energy sharing in modular self-reconfigurable robots ensures the energy balance of the modules, thus allowing the system to work sustainably. This paper proposes an energy sharing mechanism for a novel modular self-reconfigurable robot that allows free connections among modules, termed as FreeBOT, such that each FreeBOT can share energy with peers through surface contact. Corresponding energy sharing rules are proposed to achieve an energy sharing network structure without invalid components. As alternative choices, several types of networks subjected to the above requirements are provided, which also maximize the number of FreeBOTs joining to share energy. We implement and test the prototype of the energy sharing mechanism on FreeBOT. The experimental results show that the mechanism can effectively achieve energy sharing among FreeBOTs.

This paper introduces a cube-based reconfigurable robot that utilizes an electromagnet-based actuation framework to reconfigure in three dimensions via pivoting. While a variety of actuation mechanisms for self-reconfigurable robots have been explored, they often suffer from cost, complexity, assembly and sizing requirements that prevent scaled production of such robots. To address this challenge, we use an actuation mechanism based on electromagnets embedded into the edges of each cube to interchangeably create identically or oppositely polarized electromagnet pairs, resulting in repulsive or attractive forces, respectively. By leveraging attraction for hinge formation, and repulsion to drive pivoting maneuvers, we can reconfigure the robot by voxelizing it and actuating its constituent modules-termed Electrovoxels-via electromagnetically actuated pivoting. To demonstrate this, we develop fully untethered, three-dimensional self-reconfigurable robots and demonstrate 2D and 3D self-reconfiguration using pivot and traversal maneuvers on an air-table and in microgravity on a parabolic flight. This paper describes the hardware design of our robots, its pivoting framework, our reconfiguration planning software, and an evaluation of the dynamical and electrical characteristics of our system to inform the design of scalable self-reconfigurable robots.

This paper proposes Elastic Tracker, a flexible trajectory planning framework that can deal with challenging tracking tasks with guaranteed safety and visibility. Firstly, an object detection and intension-free motion prediction method is designed. Then an occlusion-aware path finding method is proposed to provide a proper topology. A smart safe flight corridor generation strategy is designed with the guiding path. An analytical occlusion cost is evaluated. Finally, an effective trajectory optimization approach enables to generate a spatio-temporal optimal trajectory within the resultant flight corridor. Particular formulations are designed to guarantee both safety and visibility, with all the above requirements optimized jointly. The experimental results show that our method works more robustly but with less computation than the existing methods, even in some challenging tracking tasks.

We propose a framework for robust and efficient training of Dense Object Nets (DON) [1] with a focus on industrial multi-object robot manipulation scenarios. DON is a popular approach to obtain dense, view-invariant object descriptors, which can be used for a multitude of downstream tasks in robot manipulation, such as, pose estimation, state representation for control, etc. However, the original work [1] focused training on singulated objects, with limited results on instance-specific, multi-object applications. Additionally, a complex data collection pipeline, including 3D reconstruction and mask annotation of each object, is required for training. In this paper, we further improve the efficacy of DON with a simplified data collection and training regime, that consistently yields higher precision and enables robust tracking of keypoints with less data requirements. In particular, we focus on training with multi-object data instead of singulated objects, combined with a well-chosen augmentation scheme. We additionally propose an alternative loss formulation to the original pixel wise formulation that offers better results and is less sensitive to hyperparameters. Finally, we demonstrate the robustness and accuracy of our proposed framework on a real-world robotic grasping task.

A key proficiency an autonomous mobile robot must have to perform high-level tasks is a strong understanding of its environment. This involves information about what types of objects are present, where they are, what their spatial extend is, and how they can be reached, i.e., information about free space is also crucial. Semantic maps are a powerful instrument providing such information. However, applying semantic segmentation and building 3D maps with high spatial resolution is challenging given limited resources on mobile robots. In this paper, we incorporate semantic information into efficient occupancy normal distribution transform (NDT) maps to enable real-time semantic mapping on mobile robots. On the publicly available dataset Hypersim, we show that, due to their sub-voxel accuracy, semantic NDT maps are superior to other approaches. We compare them to the recent state-of-the-art approach based on voxels and semantic Bayesian spatial kernel inference (S-BKI) and to an optimized version of it derived in this paper. The proposed semantic NDT maps can represent semantics to the same level of detail, while mapping is 2.7 to 17.5 times faster. For the same grid resolution, they perform significantly better, while mapping is up to more than 5 times faster. Finally, we prove the real-world applicability of semantic NDT maps with qualitative results in a domestic application.

Prehensile object rearrangement in cluttered and confined spaces has broad applications but is also challenging. For instance, rearranging products in a grocery shelf means that the robot cannot directly access all objects and has limited free space. This is harder than tabletop rearrangement where objects are easily accessible with top-down grasps, which simplifies robot-object interactions. This work focuses on problems where such interactions are critical for completing tasks. It proposes a new efficient and complete solver under general constraints for monotone instances, which can be solved by moving each object at most once. The monotone solver reasons about robot-object constraints and uses them to effectively prune the search space. The new monotone solver is integrated with a global planner to solve non-monotone instances with high-quality solutions fast. Furthermore, this work contributes an effective pre-processing tool to significantly speed up online motion planning queries for rearrangement in confined spaces. Experiments further demonstrate that the proposed monotone solver, equipped with the pre-processing tool, results in 57.3% faster computation and 3 times higher success rate than state-of-the-art methods. Similarly, the resulting global planner is computationally more efficient and has a higher success rate, while producing high-quality solutions for non-monotone instances (i.e., only 1.3 additional actions are needed on average). Videos of demonstrating solutions on a real robotic system and codes can be found at https://github.com/Rui1223/uniform_object_rearrangement.