IROS 2022 Papers

This paper presents a reduction mechanism for robot actuators that can switch between two types of reduction ratio. By fixing the carrier or ring gear of the proposed actuator which is based on the 3K compound planetary drive, the actuator can shift its reduction ratio. For compact design with reduced weight of the actuator, unique pawl brake mechanism interacting with cams and micro servos for switching mechanism is designed. The resulting prototype module has a reduction ratio of 6.91 and 44.93 for ‘low-reduction’ and ‘high-reduction’ ratios, respectively. Reduction ratios can be easily adjusted by modifying the pitch diameters of gears. Experimental results demonstrate that the proposed actuator could extend its operation region via two reduction modes that are interchangeable with gear shifting.

In this paper, we propose a multi-objective camera ISP framework that utilizes Deep Reinforcement Learning (DRL) and camera ISP toolbox that consist of network-based and conventional ISP tools. The proposed DRL-based camera ISP framework iteratively selects a proper tool from the toolbox and applies it to the image to maximize a given vision task-specific reward function. For this purpose, we implement total 51 ISP tools that include exposure correction, color-and-tone correction, white balance, sharpening, denoising, and the others. We also propose an efficient DRL network architecture that can extract the various aspects of an image and make a rigid mapping relationship between images and a large number of actions. Our proposed DRL-based ISP framework effectively improves the image quality according to each vision task such as RAW-to-RGB image restoration, 2D object detection, and monocular depth estimation.

Visual SLAM methods based on point features have achieved acceptable results in texture-rich static scenes, but they often suffer from a deficiency of texture and the existence of dynamic objects in real indoor scenes, which limits the application of these methods. In this paper, we have presented DRG-SLAM, which combines line features and plane features into point features to improve the robustness of the system. We tested the proposed algorithm on publicly available datasets, and the results demonstrate that the algorithm has superior accuracy and robustness in indoor dynamic scenes compared with the state-of-the-art methods.

An essential task for a multi-robot system is generating a common understanding of the environment and relative poses between robots. Cooperative tasks can be executed only when a vehicle has knowledge of its own state and the states of the team members. However, this has primarily been achieved with direct rendezvous between underwater robots, via inter-robot ranging. We propose a novel distributed multi-robot simultaneous localization and mapping (SLAM) framework for underwater robots using imaging sonar-based perception. By passing only scene descriptors between robots, we do not need to pass raw sensor data unless there is a likelihood of inter-robot loop closure. We utilize pairwise consistent measurement set maximization (PCM), making our system robust to erroneous loop closures. The functionality of our system is demonstrated using two real-world datasets, one with three robots and another with two robots. We show that our system effectively estimates the trajectories of the multi-robot system and keeps the bandwidth requirements of inter-robot communication low. To our knowledge, this paper describes the first instance of multi-robot SLAM using real imaging sonar data (which we implement offline, using simulated communication). Code link: https://github.com/jake3991/DRACo-SLAM.

For robots to move in the real world, they must first correctly understand the state of its own body and the tools that it holds. In this research, we propose DIJE, an algorithm to estimate the image Jacobian for every pixel. It is based on an optical flow calculation and a simplified Kalman Filter that can be efficiently run on the whole image in real time. It does not rely on markers nor knowledge of the robotic structure. We use the DIJE in a self-recognition process which can robustly distinguish between movement by the robot and by external entities, even when the motion overlaps. We also propose a visual servoing controller based on DIJE, which can learn to control the robot's body to conduct reaching movements or bimanual tool-tip control. The proposed algorithms were implemented on a physical musculoskeletal robot and its performance was verified. We believe that such global estimation of the visuomotor policy has the potential to be extended into a more general framework for manipulation.

A loop closure module plays an important role in visual SLAM systems, which can reduce the accumulat-ed drift. This task faces the challenges of large viewpoint changes and expensive computational costs when optimizing the global map. This paper proposes DH-LC, a novel accurate and robust loop closure method that consists of hierarchical spatial feature matching (HSFM) and hybrid bundle adjustment (HBA). HSFM estimates a reliable relative pose between the query image and the retrieval image in a coarse-to-fine way. Specifically, 3D points are firstly triangulated and then clus-tered according to the spatial distribution. The cluster centers estimate coarse cube-level matching pairs in a larger perception field which can tolerate large viewpoint changes. HBA optimizes the global map efficiently by adaptively selecting incremental bundle adjustment or full bundle adjustment according to the accumulated drift and relative pose verification in the temporal window. Experimental results demonstrate that our proposed method easily detects loops in large viewpoint changes and efficiently optimizes the global map. When compared with the state-of-the-art methods, our method increases loop closure recall and improves SLAM localization accuracy with reducing the accumulated drift.

This paper introduces DGBench, a fully reproducible open-source testing system to enable benchmarking of dynamic grasping in environments with unpredictable relative motion between robot and object. We use the proposed benchmark to compare several visual perception arrangements. Traditional perception systems developed for static grasping are unable to provide feedback during the final phase of a grasp due to sensor minimum range, occlusion, and a limited field of view. A multi-camera eye-in-hand perception system is presented that has advantages over commonly used camera configurations. We quantitatively evaluate the performance on a real robot with an image-based visual servoing grasp controller and show a significantly improved success rate on a dynamic grasping task.

We present a novel reinforcement learning (RL) based task allocation and decentralized navigation algorithm for mobile robots in warehouse environments. Our approach is designed for scenarios in which multiple robots are used to perform various pick up and delivery tasks. We consider the problem of joint decentralized task allocation and navigation and present a two level approach to solve it. At the higher level, we solve the task allocation by formulating it in terms of Markov Decision Processes and choosing the appropriate rewards to minimize the Total Travel Delay (TTD). At the lower level, we use a decentralized navigation scheme based on ORCA that enables each robot to perform these tasks in an independent manner, and avoid collisions with other robots and dynamic obstacles. We combine these lower and upper levels by defining rewards for the higher level as the feedback from the lower level navigation algorithm. We perform extensive evaluation in complex warehouse layouts with large number of agents and highlight the benefits over state-of-the-art algorithms based on myopic pickup distance minimization and regret-based task selection. We observe improvement up to 14% in terms of task completion time and up-to 40% improvement in terms of computing collision-free trajectories for the robots.

Multi-agent reinforcement learning (MARL) meth-ods face a curse of dimensionality in the policy and value function representations as the number of agents increases. The development of distributed or parallel training techniques is also hindered by the global coupling among the agent dynamics, requiring simultaneous state transitions. This paper introduces Distributed multi-Agent Reinforcement Learning with One-hop Neighbors (DARLIN). DARLIN is an off-policy actor-critic MARL method that breaks the curse of dimensionality and achieves distributed training by restricting the agent interactions to one-hop neighborhoods. Each agent optimizes its value and policy functions over a one-hop neighborhood, reducing the representation complexity, yet maintaining expressiveness by training with varying numbers and states of neighbors. This structure enables the key contribution of DARLIN: a distributed training procedure in which each compute node simulates the state transitions of only a small subset of the agents, greatly accelerating the training of large-scale MARL policies. Comparisons with state-of-the-art MARL methods show that DARLIN significantly reduces training time without sacrificing policy quality as the number of agents increases.

The current LiDAR SLAM (Simultaneous Localization and Mapping) system suffers greatly from low accuracy and limited robustness when faced with complicated circumstances. From our experiments, we find that current LiDAR SLAM systems have limited performance when the noise level in the obtained point clouds is large. Therefore, in this work, we propose a general framework to tackle the problem of denoising and loop closure for LiDAR SLAM in complex environments with many noises and outliers caused by reflective materials. Current approaches for point clouds denoising are mainly designed for small-scale point clouds and can not be extended to large-scale point clouds scenes. In this work, we firstly proposed a lightweight network for large-scale point clouds denoising. Subsequently, we have also designed an efficient loop closure network for place recognition in global optimization to improve the localization accuracy of the whole system. Finally, we have demonstrated by extensive experiments and benchmark studies that our method can have a significant boost on the localization accuracy of the LiDAR SLAM system when faced with noisy point clouds, with a marginal increase in computational cost.

In-hand pivoting is one of the important manipulation skills that leverage robot grippers' extrinsic dexterity to perform repositioning tasks to compensate for environmental uncertainties and imprecise motion execution. Although many researchers have been trying to solve pivoting problems using mathematical modeling or learning-based approaches, the problems remain as open challenges. On the other hand, humans perform in-hand manipulation with remarkable precision and speed. Hence, the solution could be provided by making full use of this intrinsic human skill through dexterous teleoperation. For dexterous teleoperation to be successful, interfaces that enhance and complement haptic feedback are of great necessity. In this paper, we propose a cutaneous feedback interface that complements the somatosensory information humans rely on when performing dexterous skills. The interface is designed based on five-bar link mechanisms and provides two contact points in the index finger and thumb for cutaneous feedback. By integrating the interface with a commercially available haptic device, the system can display information such as grasping force, shear force, friction, and grasped object's pose. Passive pivoting tasks inside a numerical simulator Isaac Sim is conducted to evaluate the effect of the proposed cutaneous feedback interface.

Road detection plays a fundamental role in the visual navigation system of autonomous vehicles. However, it's still challenging to achieve robust road detection in off-road scenarios due to their complicated road appearances and ambiguous road structures. Therefore, existing image-based road detection approaches usually fail to extract the right routes due to the lack of the effective fusion of the image and prior reference paths(road guidances generated via map annotations and GPS localization). Besides, the reference paths are not always reliable because of GPS localization errors and mapping errors. To achieve robust road detection in off-road scenarios, we propose a prior-correction-based road detection network named PR-ROAD via fusing the cross-model information provided by both the reference path and the input image. These two heterogeneous data, prior and image, are deeply fused by a cross-attention module and formulate contextual inter-dependencies. We conduct experiments in our collected rural, off-road and urban datasets. The experimental results demonstrate the effectiveness of the proposed method both on unstructured and structured roads.

In an attempt to nurture children's creativity, we developed a creative conversational agent to be used in a collaborative storytelling context with a child. We presented a novel approach to develop creative Artificial Intelligence (AI). Our approach uses the four creativity measures: fluency, flexi-bility, elaboration and originality in order to generate creative behavior. We analyzed and annotated our previously collected storytelling data sets -collected with children- according to our four creativity measures. We then used the extracted and annotated data (636 statements) in order to fine-tune two pre-trained language models (Open AI GPT-3). The two models were aimed at generating creative versus non-creative behavior in a collaborative storytelling scenario. We developed the two models to be able to assess the results and compare them together. We conducted an evaluation to assess stories generated collaboratively between a human and both agents separately (n = 26). Adult Users rated the creativity of the agent according to the stories generated. Results showed that the creative agent was perceived as significantly more creative than the non-creative agent. With the experiment results confirming the validity of our system, we may therefore proceed with testing the effects of the creative behavior of the agent on children's creativity skills.

Laboratory automation is a suitable solution to establish higher reproducibility with less manual work and thus higher quality standards in life sciences. To date, mobile robots are capable of performing autonomous pick-and-place tasks in the laboratory, and specialized pipetting machines can be used for sequenced liquid handling. However, the complex and creative process of developing new research protocols requires flexible robotic systems that can perform tasks such as pipetting in more versatile ways. In addition, the correct technique, according to ISO standards, has a great influence on precision and accuracy and therefore on reproducibility. This paper introduces our Intelligent Robotic Lab Assistants in the framework of our holistic, human-like, but standardized paradigm for collaborative lab automation, AI.Laboratory. Our system demonstrates mastery of pipetting following ISO 8655 as a force-sensitive robotic manipulation skill, which is a key component of our taxonomy of cell culture skills and the first steps toward true intelligent robotic laboratory assistants. This intelligent robotic pipetting skill is a versatile tool for general handling of µL-liquids, using only standard laboratory equipment that can be flexibly positioned in the robot's workspace. To demonstrate its pipetting performance, flexible handling of small volumes from 10 µL to 1000 µL was experimentally validated to the ISO 8655 standard, demonstrating superhuman performance that outperformed laymen, human experts, and other commercial and non-commercial robotic pipetting systems.

Human and robot partners increasingly need to work together to perform tasks as a team. Robots designed for such collaboration must reason about how their task-completion strategies interplay with the behavior and skills of their human team members as they coordinate on achieving joint goals. Our goal in this work is to develop a computational framework for robot adaptation to human partners in human-robot team collaborations. We first present an algorithm for autonomously recognizing available task-completion strategies by observing human-human teams performing a collaborative task. By transforming team actions into low dimensional representations using hidden Markov models, we can identify strategies without prior knowledge. Robot policies are learned on each of the identified strategies to construct a Mixture-of-Experts model that adapts to the task strategies of unseen human partners. We evaluate our model on a collaborative cooking task using an Overcooked simulator. Results of an online user study with 125 participants demonstrate that our framework improves the task performance and collaborative fluency of human-agent teams, as compared to state of the art reinforcement learning methods.

We present a new algorithm for coordinating the motion of multiple extruders to increase throughput in fused filament fabrication (FFF)/fused deposition modeling (FDM) additive manufacturing. Platforms based on FFF are commonly available and advantageous to several industries, but are limited by slow fabrication time and could be could be significantly improved through efficient use of multiple extruders. We propose the coordinated toolpath planning problem for systems of extruders mounted as end-effectors on robot arms with the objective of maximizing utilization and avoiding collisions. Building on the idea of dependency graphs introduced in our earlier work, we develop a planning and control framework that precomputes a set of multi-layer toolpath segments from the input model and efficiently assigns them to individual extruders such that executed toolpaths are collision-free. Our method overcomes key limitations of existing methods, including utilization loss from workspace partitioning, precomputed toolpaths subject to collisions with the partially fabricated object, and wasted motion resulting from strict layer-by-layer fabrication. We report simulation results that show a major increase in utilization compared to single and multi-extruder methods, and favorable fabrication results using commodity hardware that demonstrate the feasibility of our method in practice.

The lack of communication between agents in a multi-robot system is often regarded as a limiting factor that can affect and delay cooperative exploration and exploitation of cluttered and uncertain environments. On the contrary, this paper proposes a complete planning framework to enable cooperative behavior without the need for constant communication between robots, demonstrating drastic improvements in task completion and coverage time as compared to both fully connected robotic networks and widely used frontier-based exploration methods. Specifically, the proposed scheme considers three behaviors: i) exploration, promoting separation and disconnection, ii) rendezvous to reconnect and share information gained during exploration, and iii) task allocation for prioritized objectives. Exploration is achieved via a Sobel edge detection frontier algorithm that enables navigation of unknown complex (both convex and non-convex) environments. Once a task is discovered, a multi-objective weighted sum optimization method is proposed for allocating tasks based on prioritization and expectation estimation. The utility, generality, and scalability of the proposed approach is demonstrated using extensive simulations and experiments with unmanned ground vehicles in various cluttered environments.

As a type of body language, gestures can largely affect the impressions of human-like robots perceived by users. Recent data-driven approaches to the generation of co-speech gestures have successfully promoted the naturalness of produced gestures. These approaches also possess greater generalizability to work under various contexts than rule-based methods. However, most have no direct control over the human impressions of robots. The main obstacle is that creating a dataset that covers various impression labels is not trivial. In this study, based on previous findings in cognitive science on robot impressions, we present a heuristic method to control them without manual labeling, and demonstrate its effectiveness on a virtual agent and partially on a humanoid robot through subjective experiments with 50 participants.

Dynamic movements are ubiquitous in human motor behavior as they tend to be more efficient and can solve a broader range of skill domains than their quasi-static counterparts. For decades, robotic juggling tasks have been among the most frequently studied dynamic manipulation problems since the required dynamic dexterity can be scaled to arbitrarily high difficulty. However, successful approaches have been limited to basic juggling skills, indicating a lack of understanding of the required constraints for dexterous toss juggling. We present a detailed analysis of the toss juggling task, identifying the key challenges of the switching contacts task and formalizing it as a trajectory optimization problem. Building on our state-of-the-art, real-world toss juggling platform, we reach the theoretical limits of toss juggling in simulation, evaluate a resulting real-time controller in environments of varying difficulty and achieve robust toss juggling of up to 17 balls on two anthropomorphic manipulators. https://sites.google.com/view/controlling-the-cascade