IROS 2022 论文列表

Sampling-based motion planners are widely used in robotics due to their simplicity, flexibility and computational efficiency. However, in their most basic form, these algorithms operate under the assumption of static scenes and lack the ability to avoid collisions with dynamic (i.e. moving) obstacles. This raises safety concerns, limiting the range of possible applications of mobile robots in the real world. Motivated by these challenges, in this work we present Temporal-PRM, a novel sampling-based path-planning algorithm that performs obstacle avoidance in dynamic environments. The proposed approach extends the original Probabilistic Roadmap (PRM) with the notion of time, generating an augmented graph-like structure that can be efficiently queried using a time-aware variant of the A* search algorithm, also introduced in this paper. Our design maintains all the properties of PRM, such as the ability to perform multiple queries and to find smooth paths, while circumventing its downside by enabling collision avoidance in highly dynamic scenes with a minor increase in the computational cost. Through a series of challenging experiments in highly cluttered and dynamic environments, we demonstrate that the proposed path planner outperforms other state-of-the-art sampling-based solvers. Moreover, we show that our algorithm can run onboard a flying robot, performing obstacle avoidance in real time.

This paper focuses on problems associated with the deployment of automatic agents for last-mile delivery. We propose a framework and methodology to systematically evaluate and compare different hybrid strategies. Performance metrics in agent noise, delivery time, energy consumption, coverage rate, package throughput, and system costs are defined rigorously and modeled mathematically. Using the methodology, we conduct a case study in the city of Boston for four agent delivery strategies, including a hybrid strategy proposed in this paper. The proposed strategy utilizes available space in public transits' cabins during off-peak hours to relocate the agent traveling start locations. Simulations and analyses show that hybrid strategies outperform the Agent-Only delivery strategy in terms of noise exposure, energy consumption, and coverage rate. The performance of hybrid strategies highly depends on the characteristics of the ground transportation methods accompanying agents. Thus, the methods of ground transportation should carefully be examined and selected for each case and strategy in real-world applications.

In physical human-robot interaction (pHRI) enabled by admittance control, delay-induced oscillations arising from both the neuromuscular time-delays of the human and electromechanical delays of the robot can cause unsafe instability in the system. This study presents and evaluates rate-limiting as a means to overcome such instability, and provides a new perspective on how rate-limiting can benefit pHRI. Specifically, a rate-limited and time-delayed human-in-the-loop (HITL) model is analyzed to show not only how the rate-limiter can transform an unstable equilibrium (due to time-delay) into a stable limit-cycle, but also how a desired upper-bound on the range of persistent oscillations can be achieved by appropriately setting the rate-limiter threshold. In addition, a study involving 10 subjects and the EXO-UL8 upper-limb exoskeleton, and consisting of 16 trials - 4 rate-limiter thresholds by 4 time-delays - is performed to: (1) validate the relationships between time-delays, rate-limits, and position bounds on persistent oscillations, and (2) demonstrate the effectiveness of rate-limiting for recovery from delay-induced oscillations without interfering with regular operation. Agreement of experimental results with the theoretical developments supports the feasibility of incorporating rate-limiting in admittance-controlled pHRI systems as a safety mechanism.

Autonomous agents need to perceive the world in a robust way, such that the shift in data distribution does not lead to faulty perception results. When agents cannot be trained with abundant data, agents may need to operate on real world environments while trained on simulated data, and suffer from domain shift. This paper proposes an effective and robust unsupervised domain adaptation (UDA) method that can resolve these situations. In the UDA setup, we are given a labeled source domain and an unlabeled target domain that share the same set of classes but are sampled from different distributions. This domain shift prevents agents which employ deep neural networks from generalizing well on the target domain. Recent methods adopt the strategy of self-training the networks with pseudo labeled target samples. However, falsely labeled samples cause negative transfer and deteriorate generalization of a network. to reduce negative transfer we propose an algorithm that can filter the pseudo labels, and use the filtered labels to align the domains in the feature space. The samples whose labels have not passed the filtering process can be used as an index to tune the hyperparameters of our method. Across various benchmarks, we validate the performance of our method. Especially, our method achieves strong performance on the synthetic-to-real adaptation scenario.

We present a strong baseline that surpasses the performance of previously published methods on the Habitat Challenge task of navigating to a target object in indoor environments. Our method is motivated from primary failure modes of prior state-of-the-art: poor exploration, inaccurate object identification, and agent getting trapped due to imprecise map construction. We make three contributions to mitigate these issues: (i) First, we show that existing map-based methods fail to effectively use semantic clues for exploration. We present a semantic-agnostic exploration strategy (called Stubborn) without any learning that surprisingly outperforms prior work. (ii) We propose a strategy for integrating temporal information to improve object identification. (iii) Lastly, due to inaccurate depth observation the agent often gets trapped in small regions. We develop a multi-scale collision map for obstacle identification that mitigates this issue. Website: https://github.com/Improbable-AI/Stubborn

We model a finite horizon decision making process between an ego and a non-ego vehicle, where the non-ego vehicle has a certain probability of moving adversarially over each planning stage. The adversarial intent of the non-ego vehicle is inferred only when a particular set of actions are performed by both vehicles, thereby creating a stopping state. We term such a decision-making process as a multi-stage stochastic zero-sum game (SSG) with stopping states, i.e., once adversarial intent is ascertained, the non-ego vehicle continues to choose its actions adversarially for the remaining stages of the interaction. We analytically characterize the Nash equilibria of this game for the case of two actions per player. We then demonstrate this approach via two autonomous motion planning applications. The first involves maintaining a safe distance from a non-ego vehicle ahead, modeled using fixed stage costs. The second involves safe lane-changing with costs that are stage dependent. In both scenarios, we provide a comparison between the analytic/simulated and experimental results using ground robots.

Keeping balance is one of the most important tasks of a motorcycle. The steady-state manifold is proposed in this paper to explore the inherent dynamics and the balance properties of a riderless motorcycle. The dynamic and kinematic characteristics are analyzed based on the manifold and are validated by simulation. Comparing to traditional control method, the usefulness of the manifold in control is shown through the design of a novel control strategy. Furthermore, based on the analysis and the simulation, the potential applications of the manifold for control and planning are summarized.

Traversing intersections is a challenging problem for autonomous vehicles, especially when the intersections do not have traffic control. Recently deep reinforcement learning has received massive attention due to its success in dealing with autonomous driving tasks. In this work, we address the problem of traversing unsignalized intersections using a novel curriculum for deep reinforcement learning. The proposed curriculum leads to: 1) A faster training process for the reinforcement learning agent, and 2) Better performance compared to an agent trained without curriculum. Our main contribution is two-fold: 1) Presenting a unique curriculum for training deep reinforcement learning agents, and 2) demonstrating the performance improvement using the proposed curriculum in the unsignalized intersection traversal task. The framework expects processed observations of the surroundings from the perception system of the autonomous vehicle. We test our method in the CommonRoad motion planning simulator on T-intersections and four-way intersections.

Force exertion, object manipulation, and interaction are novel trending research topics of autonomous flying robots that can yield hoovering. Moreover, specifically with quadrotors, the vibration caused by the high natural frequency of rotating propellers exacerbates the problem of maintaining contact and exerting force against a rigidly fixed object. This contact vibration transfers back kinetic energy to the quadrotor that, in worst-case scenarios, surpasses its flying capabilities, which may lead to a crash. This paper studies the problem of aerial contact stabilization of a quadrotor equipped with a hemispherical deformable tip, which accommodates contact forces at a lower frequency. Thus two phenomena not studied in the literature arise: the rolling motion, and the deformation at contact. The contact force stabilization restores the effects of deformation while simultaneously endowing rolling by controlling a tangent constrained force. A model-free continuous attitude fractional controller to guarantee finite-time attitude stabilization is proposed. The residual coupled nonlinear dynamics yields the desired attitude corresponding to a given contact force; thus, force stabilization is achieved. Finally, experimental results are presented to assess the performance of the proposed approach.

Sound source separation is a method to extract a target sound source from a mixture of various sound sources and noises. One of the typical sound source separation methods is beamforming, which can separate sound sources by direction based on the phase difference between channels from the recorded signal of a microphone array, a multi-channel recording system. However, beamforming is a direction-based method and cannot separate multiple sources in the same direction. In this paper, we propose a method for separating sources in the same direction using multiple microphone arrays. The proposed method performs beamforming using multiple microphone arrays and extracts only the target sound source from the separated sound by the Non-negative Matrix Factorization (NMF), thus reducing the influence of other sources in the same direction. In this paper, to investigate the effectiveness of the proposed method, experiments were conducted assuming the presence of another sound source in the same direction from an arbitrary microphone array. The results show that the proposed method outperforms the delay-sum method in a simulation environment. In addition, experiments were conducted in a real environment to verify the effect of reverberation.

In this paper, we consider online planning in par-tially observable domains. Solving the corresponding POMDP problem is a very challenging task, particularly in an online setting. Our key contribution is a novel algorithmic approach, Simplified Information Theoretic Belief Space Planning (SITH-BSP), which aims to speed up POMDP planning considering belief-dependent rewards, without compromising the solution's accuracy. We do so by mathematically relating the simplified el-ements of the problem to the corresponding counterparts of the original problem. Specifically, we focus on belief simplification and use it to formulate bounds on the corresponding original belief-dependent rewards. These bounds in turn are used to perform branch pruning over the belief tree, in the process of extracting the optimal policy from this existing belief tree. We further introduce the notion of adaptive simplification, while re-using calculations between different simplification levels, and exploit it to prune, at each level in the belief tree, all branches but one. Therefore, our approach is guaranteed to find the optimal solution (policy) that corresponds to the given belief tree but with substantial speedup. As a second key contribution, we derive novel analytical bounds for differential entropy, considering a sampling-based belief representation, which we believe are of interest on their own. We validate our approach in simulation using these bounds and where simplification corresponds to reducing the number of samples, exhibiting a significant computational speedup while yielding the optimal solution for the given belief tree.

Planning collision-free motions for robots with many degrees of freedom is challenging in environments with complex obstacle geometries. Recent work introduced the idea of speeding up the planning by encoding prior experience of successful motion plans in a neural network. However, this “neural motion planning” did not scale to complex robots in unseen 3D environments as needed for real-world applications. Here, we introduce “basis point set”, well-known in computer vision, to neural motion planning as a modern compact environment encoding enabling efficient supervised training networks that generalize well over diverse 3D worlds. Combined with a new elaborate training scheme, we reach a planning success rate of 100 %. We use the network to predict an educated initial guess for an optimization-based planner (OMP), which quickly converges to a feasible solution, massively outperforming random multi-starts when tested on previously unseen environments. For the DLR humanoid Agile Justin with 19 DoF and in challenging obstacle environments, optimal paths can be generated in 200 ms using only a single CPU core. We also show a first successful real-world experiment based on a high-resolution world model from an integrated 3D sensor.

Folding garments reliably and efficiently is a long standing challenge in robotic manipulation due to the complex dynamics and high dimensional configuration space of garments. An intuitive approach is to initially manipulate the garment to a canonical smooth configuration before folding. In this work, we develop SpeedFolding, a reliable and efficient bimanual system, which given user-defined instructions as folding lines, manipulates an initially crumpled garment to (1) a smoothed and (2) a folded configuration. Our primary contribution is a novel neural network architecture that is able to predict pairs of gripper poses to parameterize a diverse set of bimanual action primitives. After learning from 4300 human- annotated and self-supervised actions, the robot is able to fold garments from a random initial configuration in under 120 s on average with a success rate of 93 %. Real-world experiments show that the system is able to generalize to unseen garments of different color, shape, and stiffness. While prior work achieved 3–6 Folds Per Hour (FPH), SpeedFolding achieves 30–40 FPH. See https://pantor.github.io/speedfolding for code, videos, and datasets.

For wave-driven unmanned surface vehicles (WUSVs), utilizing oscillating foils is the most straightforward and common wave energy conversion mechanism. Improving the thrust of the oscillating foil to increase its speed can help WUSVs improve their maneuverability and shorten the completion of ocean missions. This paper proposes a novel transformable two-segment foil, improving the wave energy-converting efficiency to provide more average thrust in every wave cycle. We estimate their working effectiveness numerically with a simple model to verify that the design enhances foils' thrust force. The thrust enhancement was further confirmed by computational fluid dynamic (CFD) simulations, and we estimated the suitable values of parameters of the foils in several different common sea conditions in coastal waters by CFD simulations. We design and make two wave gliders with traditional and transformable two-segment foils and finish the speed enhancement experiments. The speed enhancement is verified, and transformable two-segment foils can increase the speed of WUSVs by 10% in similar sea conditions in experiments.

Simultaneous localization and mapping (SLAM) is a critical capability in autonomous navigation, but in order to scale SLAM to the setting of “lifelong” SLAM, particularly under memory or computation constraints, a robot must be able to determine what information should be retained and what can safely be forgotten. In graph-based SLAM, the number of edges (measurements) in a pose graph determines both the memory requirements of storing a robot's observations and the computational expense of algorithms deployed for performing state estimation using those observations; both of which can grow unbounded during long-term navigation. To address this, we propose a spectral approach for pose graph sparsification which maximizes the algebraic connectivity of the sparsified measurement graphs, a key quantity which has been shown to control the estimation error of pose graph SLAM solutions. Our algorithm, MAC (for maximizing algebraic connectivity), which is based on convex relaxation, is simple and computationally inexpensive, and admits formal post hoc performance guarantees on the quality of the solutions it provides. In experiments on benchmark pose-graph SLAM datasets, we show that our approach quickly produces high-quality sparsification results which retain the connectivity of the graph and, in turn, the quality of corresponding SLAM solutions, as compared to a baseline approach which does not consider graph connectivity.

Recovering depth information from a single image is a long-standing challenge, and self-supervised depth estimation methods have gradually attracted attention due to not relying on high-cost ground truth. Constructing an accurate photometric loss based on photometric consistency is crucial for these self-supervised methods to obtain high-quality depth maps. However, the photometric loss in most studies treats all pixels indiscriminately, resulting in poor performance. In this paper, we propose two modules based on the spatial and temporal cues to refine the photometric loss. Delving into the geometric model of photometric consistency, we introduce a depth-aware pixel correspondence module (DPC) inside the monocular depth estimation pipeline. It reduces the uncertainty of photometric errors by applying the homography matrix to the projection of corresponding pixels in far regions instead of the fundamental matrix. Furthermore, we design an omnidirectional auto-masking module (OA) to boost the robustness of our model, which utilizes temporal sequences to generate disturbance poses and hypothetical views to distin-guish dynamic objects with different directions that violate the photometric consistency. Experiments on the KITTI and the Make3d datasets reveal that our framework achieves state-of-the-art performance.

Localization in topological maps is essential for image-based navigation using an RGB camera. Localization using only one camera can be challenging in medium-to-large-sized environments because similar-looking images are often observed repeatedly, especially in indoor environments. To overcome this issue, we propose a learning-based localization method that simultaneously utilizes the spatial consistency from topological maps and the temporal consistency from time-series images captured by a robot. Our method combines a convolutional neural network (CNN) to embed image features and a recurrent-type graph neural network to perform accurate localization. When training our model, it is difficult to obtain the ground truth (GT) pose of the robot when capturing images in real-world environments. Hence, we propose a sim2real transfer approach with semi-supervised learning that leverages simulator images with the GT pose in addition to real images. We evaluated the proposed method quantitatively and qualitatively and compared it with several state-of-the-art baselines. The proposed method outperformed the baselines in environments where the map contained similar images. Moreover, we evaluated an image-based navigation system incorporating our localization method and confirmed that navigation accuracy significantly improved in the simulator and real environments compared to the other baseline methods.

Bird's Eye View (BEV) is a popular representation for processing 3D point clouds, and by its nature is fundamentally sparse. Motivated by the computational limitations of mobile robot platforms, we create a fast, high-performance BEV 3D object detector that maintains and exploits this input sparsity to decrease runtimes over non-sparse baselines and avoids the tradeoff between pseudoimage area and runtime. We present results on KITTI, a canonical 3D detection dataset, and Matterport-Chair, a novel Matterport3D-derived chair detection dataset from scenes in real furnished homes. We evaluate runtime characteristics using a desktop GPU, an embedded ML accelerator, and a robot CPU, demonstrating that our method results in significant detection speedups (2 × or more) for embedded systems with only a modest decrease in detection quality. Our work represents a new approach for practitioners to optimize models for embedded systems by maintaining and exploiting input sparsity throughout their entire pipeline to reduce runtime and resource usage while preserving detection performance. All models, weights, experimental configurations, and datasets used are publicly available11https://vedder.io/sparse_point_pillars.