ICRA 2023 论文列表

Safety has been a key goal for autonomous driving since its inception, and we believe recognizing and responding to risk is a key component of safety. In this work, we aim to answer the question, “How can explainable risk representations be generated and used to produce risk-averse trajectories?” To answer this question, previous work uses risk metrics to formulate an optimization problem. In contrast, our work is based on research showing the usefulness of grids as a representation to generate image-based risk maps through a trained neural network. We propose a method of determining risk from a bird's eye view (BEV) of an autonomous vehicle's surroundings. Our method consists of (1) a risk map generator, which is trained to recognize risk associated with nearby agents and the map, (2) differentiable value iteration using the risk map to learn a policy, and (3) a trajectory sampler, which samples from this policy to generate a trajectory. We evaluate our planner in a close-loop manner and find improvements in its overall ability to mimic human driving while maintaining comparable safety statistics. Self-ablation also reveals the potential for fine-tuning the behavior of the planner given a designer's needs.

In this paper, we present a novel Model Predictive Control method for autonomous robot planning and control subject to arbitrary forms of uncertainty. The proposed Risk-Aware Model Predictive Path Integral (RA-MPPI) control utilizes the Conditional Value-at-Risk (CVaR) measure to generate optimal control actions for safety-critical robotic applications. Different from most existing Stochastic MPCs and CVaR optimization methods that linearize the original dynamics and formulate control tasks as convex programs, the proposed method directly uses the original dynamics without restricting the form of the cost functions or the noise. We apply the novel RA-MPPI controller to an autonomous vehicle to perform aggressive driving maneuvers in cluttered environments. Our simulations and experiments show that the proposed RA-MPPI controller can achieve similar lap times with the baseline MPPI controller while encountering significantly fewer collisions. The proposed controller performs online computation at an update frequency of up to 80 Hz, utilizing modern Graphics Processing Units (GPUs) to multi-thread the generation of trajectories as well as the CVaR values.

We propose a compact pipeline to unify all the steps of Visual Localization: image retrieval, candidate re-ranking and initial pose estimation, and camera pose refinement. Our key assumption is that the deep features used for these individual tasks share common characteristics, so we should reuse them in all the procedures of the pipeline. Our DRAN (Deep Retrieval and image Alignment Network) is able to extract global descriptors for efficient image retrieval, use intermediate hierarchical features to re-rank the retrieval list and produce an initial pose guess, which is finally refined by means of a feature-metric optimization based on learned deep multi-scale dense features. DRAN is the first single network able to produce the features for the three steps of visual localization. DRAN achieves competitive performance in terms of robustness and accuracy under challenging conditions in public benchmarks, outperforming other unified approaches and consuming lower computational and memory cost than its counterparts using multiple networks. Code and models will be publicly available at github.com/jmorlana/DRAN.

This work proposes a robot task planning framework for retrieving a target object in a confined workspace among multiple stacked objects that obstruct the target. The robot can use prehensile picking and in-workspace placing actions. The method assumes access to 3D models for the visible objects in the scene. The key contribution is in achieving desirable properties, i.e., to provide (a) safety, by avoiding collisions with sensed obstacles, objects, and occluded regions, and (b) resolution completeness (RC) - or probabilistic completeness (PC) depending on implementation - which indicates a solution will be eventually found (if it exists) as the resolution of algorithmic parameters increases. A heuristic variant of the basic RC algorithm is also proposed to solve the task more efficiently while retaining the desirable properties. Simulation results compare using random picking and placing operations against the basic RC algorithm that reasons about object dependency as well as its heuristic variant. The success rate is higher for the RC approaches given the same amount of time. The heuristic variant is able to solve the problem even more efficiently than the basic approach. The integration of the RC algorithm with perception, where an RGB-D sensor detects the objects as they are being moved, enables real robot demonstrations of safely retrieving target objects from a cluttered shelf.

Robot-guided vascular access has the potential to deliver urgent medical care in situations where medical personnel are unavailable. However, this technique requires accurate and reliable segmentation of anatomical landmarks in the body. For the ultrasound imaging modality, obtaining large amounts of training data for a segmentation model is time-consuming and expensive. This paper introduces RESUS (RESlicing of UltraSound Images), a weak supervision data augmentation technique for ultrasound images based on slicing reconstructed 3D volumes from tracked 2D images. This technique allows us to generate views which cannot be easily obtained in vivo due to physical constraints of ultrasound imaging, and use these augmented ultrasound images to train a semantic segmentation model. We demonstrate that RESUS achieves statistically significant improvement over training with non-augmented images and highlight qualitative improvements through vessel reconstruction.

Micro aerial vehicles (MAVs) hold the potential for performing autonomous and contactless land surveys for the detection of landmines and explosive remnants of war (ERW). Metal detectors are the standard detection tool but must be operated close to and parallel to the terrain. A successful combination of MAVs with metal detectors has not been presented yet, as it requires advanced flight capabilities. To this end, we present an autonomous system to survey challenging undulated terrain using a metal detector mounted on a 5 degrees of freedom (DOF) MAV. Based on an online estimate of the terrain, our receding-horizon planner efficiently covers the area, aligning the detector to the surface while considering the kinematic and visibility constraints of the platform. As the survey requires resilient and accurate localization in diverse terrain, we also propose a factor graph-based online fusion of GNSS, IMU, and LiDAR measurements. We validate the robustness of the solution to individual sensor degeneracy by flying under the canopy of trees and over featureless fields. A simulated ablation study shows that the proposed planner reduces coverage duration and improves trajectory smoothness. Real-world flight experiments showcase autonomous mapping of buried metallic objects in undulated and obstructed terrain.

Autonomous operation in underwater environ-ments is, arguably, one of the most complex domains. It requires safe operations under the presence of unpredictable surge, currents, uncertainty, and dynamic obstacles that challenges to the highest degree real-time motion planning; the primary focus of this paper. Although previous work addressed the problem of safe real-time 3D navigation in cluttered underwater environments, it did not account explicitly for disturbances, currents, dynamic obstacles, or uncertainty growth. This paper presents ResiPlan, a novel motion planning framework that utilizes past information of errors monitoring the path follower's performance, along with estimation of dynamic obstacles and uncertainty, to produce adaptive paths by adjusting the safety margins accordingly. Extensive numerical experiments and simulations validate the safety guarantees of the technique, in a variety of different environments with various types of disturbance, showcasing the strong potential to be utilized for operations in challenging underwater environments.

Synthetic fiber ropes are widely used for robots because of their advantages such as lightweight, high tensile strength and flexibility. However, there is limited information on the physical properties of synthetic fiber ropes when used for robots. This study focuses on repetitive twisting of synthetic fiber ropes and provides information for selecting them for robots based on durability. To this end, we conducted repetitive twisting experiments on five types of ropes made from different fibers; we revealed that Dyneema has higher durability against repetitive twisting than the other ropes when a single rope is twisted. In addition, we conducted experiments on Dyneema by applying torsion to two ropes in parallel like a twisted string actuator. The result indicated that two Dyneema ropes in parallel have higher durability than a single rope; however, we revealed that the the tensile strength decreases sharply with an increase in the angle of twist.

Due to the limited cargo/functional element loading and other capabilities of individual microrobots, assembling them for locomotion and disassembling them as arriving at the target is more effective. An approach called rendezvous and docking is proposed in this paper to control the assembly and disassembly of helical microrobots actuated by a uniform rotating magnetic field. Docking is realized around the intersection of their arc orbits with the assistance of a fluidic field. To adjust the distance between the adjacent helical robots suspending in solution but with a distance beyond the acceptable range of the magnetic interaction, their asynchronized velocities are achieved using the interaction between the robots and fluids. For robots rotating at different speeds around their longitudinal axes at a driving frequency lower than the cut-off frequency, different fluidic flows will be generated. Based on the interaction between the robots and the fluids, the translational trajectory paths may be tuned, causing the adjacent robots to move closer. Docking along the tangential direction of rendezvous arc trajectories avoids the instability of the helical robot rotating around the radial direction and the problem of excessive linear speed at the end during assembly so that the robot can rotate stably around its axis while completing the assembly. Besides these, assembled microrobots can also lower the requirements on the imaging resolution of motion tracking and the forces for driving; hence much lower cost for both imaging and driving equipment.

In this work, we address a visbility-based target tracking problem in a polygonal environment in which a group of mobile observers try to maintain a line-of-sight with a mobile intruder. We build a bridge between data mining and visibility-based tracking using a novel tiling scheme for the polygon. First, we propose a tracking strategy for a team of guards located on the tiles to dynamically track an intruder when complete coverage of the polygon cannot be ensured. Next, we propose a novel variant of the Voronoi Diagram to construct navigation strategies for a team of co-located guards to track an intruder from any initial position in the environment. We present empirical analysis to illustrate the efficacy of the proposed tiling scheme. Simulations and testbed demonstrations are present in a video attachment.

In this paper, a novel method based on Artificial Potential Field (APF) theory is presented, for optimal motion planning in fully-known, static workspaces, for multiple final goal configurations. Optimization is achieved through a Reinforcement Learning (RL) framework. More specifically, the parameters of the underlying potential field are adjusted through a policy gradient algorithm in order to minimize a cost function. The main novelty of the proposed scheme lies in the method that provides optimal policies for multiple final positions, in contrast to most existing methodologies that consider a single final configuration. An assessment of the optimality of our results is conducted by comparing our novel motion planning scheme against a RRT* method.

Prior work has looked at applying reinforcement learning (RL) approaches to autonomous driving scenarios, but the safety of the algorithm is often compromised due to instability or the presence of ill-defined reward functions. With the use of control barrier functions embedded into the RL policy, we arrive at safe policies to optimize the performance of the autonomous driving vehicle through the advantage of a safety layer over the RL methods to ease the design of reward functions. However, control barrier functions need a good approximation of the model of the system. We use probabilistic control barrier functions [4] to account for model uncertainty. Our Safety-Assured Policy Optimization - Ramp Merging (SAPO-RM) algorithm is implemented online in the CARLA [1] Simulator and offline on the US I-80 dataset extracted from the NGSIM Database provided by NHTSA [2]. We further test the algorithm and perform ablation studies of it on the US-101 and exi-D datasets to compare the approaches. The proposed algorithm can also be applied to other driving scenarios by changing the reward and safety constraints.

Reinforcement learning (RL) exhibits impressive performance when managing complicated control tasks for robots. However, its wide application to physical robots is limited by the absence of strong safety guarantees. To overcome this challenge, this paper explores the control Lyapunov barrier function (CLBF) to analyze the safety and reachability solely based on data without explicitly employing a dynamic model. We also proposed the Lyapunov barrier actor-critic (LBAC), a model-free RL algorithm, to search for a controller that satisfies the data-based approximation of the safety and reachability conditions. The proposed approach is demonstrated through simulation and real-world robot control experiments, i.e., a 2D quadrotor navigation task. The experimental findings reveal this approach's effectiveness in reachability and safety, surpassing other model-free RL methods.

In this paper, we propose a method for reinforcement learning-based laser welding control. Conventional methods apply standard reinforcement learning formulations to welding tasks, but we show that this formulation can minimize bead width or penetration depth errors only when the welding speed is constant. Therefore, conventional methods are suboptimal for training control parameters including the welding speed. The proposed method discounts future rewards with respect to the welding length instead of time steps to solve this issue. This is easily implemented by (1) modifying the discount factor used for QQ-function updates in existing reinforcement learning algorithms and (2) using an appropriate reward function. Experimental results using simulators show that the proposed method achieves performance that is superior to conventional methods.

Cable driven parallel robots (CDPRs) are often challenging to model and to dynamically control due to the inherent flexibility and elasticity of the cables. The additional inclusion of online geometric reconfigurability to a CDPR results in a complex underdetermined system with highly non-linear dynamics. The necessary (numerical) redundancy resolution requires multiple layers of optimization rendering its application computationally prohibitive for real-time control. Here, deep reinforcement learning approaches can offer a model-free framework to overcome these challenges and can provide a real-time capable dynamic control. This study discusses three settings for a model-free DRL implementation in dynamic trajectory tracking: (i) for a standard non-redundant CDPR with a fixed workspace; (ii) in an end-to-end setting with redundancy resolution on a reconfigurable CDPR; and (iii) in a decoupled approach resolving kinematic and actuation redundancies individually.

Grasping an object when it is in an ungraspable pose is a challenging task, such as books or other large flat objects placed horizontally on a table. Inspired by human manipulation, we address this problem by pushing the object to the edge of the table and then grasping it from the hanging part. In this paper, we develop a model-free Deep Reinforcement Learning framework to synergize pushing and grasping actions. We first pre-train a Variational Autoencoder to extract high-dimensional features of input scenario images. One Proximal Policy Optimization algorithm with the common reward and sharing layers of Actor-Critic is employed to learn both pushing and grasping actions with high data efficiency. Experiments show that our one network policy can converge 2.5 times faster than the policy using two parallel networks. Moreover, the experiments on unseen objects show that our policy can generalize to the challenging case of objects with curved surfaces and off-center irregularly shaped objects. Lastly, our policy can be transferred to a real robot without fine-tuning by using CycleGAN for domain adaption and outperforms the push-to-wall baseline.

3D face reconstruction plays a major role in many human-robot interaction systems, from automatic face authentication to human-computer interface-based entertainment. To improve robustness against occlusions and noise, 3D face reconstruction networks are often trained on a set of in-the-wild face images preferably captured along different viewpoints of the subject. However, collecting the required large amounts of 3D annotated face data is expensive and time-consuming. To address the high annotation cost and due to the importance of training on a useful set, we propose an Active Learning (AL) framework that actively selects the most informative and representative samples to be labeled. To the best of our knowledge, this paper is the first work on tackling active learning for 3D face reconstruction to enable a label-efficient training strategy. In particular, we propose a Reinforcement Active Learning approach in conjunction with a clustering-based pooling strategy to select informative view-points of the subjects. Experimental results on 300W-LP and AFLW2000 datasets demonstrate that our proposed method is able to 1) efficiently select the most influencing view-points for labeling and outperforms several baseline AL techniques and 2) further improve the performance of a 3D Face Reconstruction network trained on the full dataset.

Reconstructing objects from real world data and rendering them at novel views is critical to bringing realism, diversity and scale to simulation for robotics training and testing. In this work, we present NeuSim, a novel approach that estimates accurate geometry and realistic appearance from sparse in-the-wild data captured at distance and at limited viewpoints. Towards this goal, we represent the object surface as a neural signed distance function and leverage both LiDAR and camera sensor data to reconstruct smooth and accurate geometry and normals. We model the object appearance with a robust physics-inspired reflectance representation effective for in-the-wild data. Our experiments show that NeuSim has strong view synthesis performance on challenging scenarios with sparse training views. Furthermore, we showcase composing NeuSim assets into a virtual world and generating realistic multi-sensor data for evaluating self-driving perception models. The supplementary material can be found at the project website: https://waabi.ai/research/neusim/

Inflatable structures have attracted considerable research attention in many fields owing to their numerous advantages, such as being light and able to engage in interactions safely. However, in most cases, the inflatable structure can only have one stable configuration, which is undesirable for robotic arms. This study proposes a novel inflatable structure that can be easily reconfigured into multiple stable configurations, even with single-body inflation. In the proposed mechanism, the structure length can be freely adjusted, and its respective joints can be set in the desired directions to facilitate the reconfiguration of its pose. An additional advantage of the proposed mechanism is that it can withstand external forces as well as its own weight. This study analyzes and experimentally validates the shape locking and load-carrying properties of the proposed mechanism. Further, the fabrication process and design guidelines for the proposed mechanism are presented. Through a suitable demonstration, the proposed mechanism is shown to exhibit multiple stable configurations and lock its poses.