ICRA 2023 论文列表

Searching for objects is a fundamental skill for robots. As such, we expect object search to eventually become an off-the-shelf capability for robots, similar to e.g., object detection and SLAM. In contrast, however, no system for 3D object search exists that generalizes across real robots and environments. In this paper, building upon a recent theoretical framework that exploited the octree structure for representing belief in 3D, we present GenMOS (Generalized Multi-Object Search), the first general-purpose system for multi-object search (MOS) in a 3D region that is robot-independent and environment-agnostic. GenMOS takes as input point cloud observations of the local region, object detection results, and localization of the robot's view pose, and outputs a 6D viewpoint to move to through online planning. In particular, GenMOS uses point cloud observations in three ways: (1) to simulate occlusion; (2) to inform occupancy and initialize octree belief; and (3) to sample a belief-dependent graph of view positions that avoid obstacles. We evaluate our system both in simulation and on two real robot platforms. Our system enables, for example, a Boston Dynamics Spot robot to find a toy cat hidden underneath a couch in under one minute. We further integrate 3D local search with 2D global search to handle larger areas, demonstrating the resulting system in a 25m2 lobby area.

Previous research into using robots for pain man-agement has shown promise. However to date, there seems to have been little research investigating user requirements for robotic pain management devices which could be used by adults living with chronic pain, and how these might be translated into custom products. We carried out a user study comprising online surveys and interviews with people who have lived experience of chronic pain to investigate their perspectives. We had a total of 44 participants in our study. Our research revealed a preference for robotic devices for pain management which have an abstract or animal-like form, noting that contact points with the body should feel soft, warm, and light. Study participants also felt that the user should initiate the interaction and should have control of the robot, as well as the type and intensity of touch. Favored touch types included massaging, rubbing, and stroking. From the emerging requirements, given the diversity of experiences, design-related attributes identified could be used for a form-customization application, such as interactive evolutionary computation (IEC), as a means to personalize the embodiment of robotic devices. Prioritized form factors for customization through included size, weight, and feel.

This paper presents a stiffness-changeable soft finger using chain mail jamming. This finger can achieve adaptive grasping and in-hand manipulation by reshaping and exerting changeable gripping force. The jamming phenomenon happens when particles in a chamber get interlocked where confining pressure is exerted at their boundaries, which is widely used to construct mechanisms with changeable stiffness. Compared with the traditional granular media, chain mail has a lower packing fraction and provides a stronger tensile force. In this paper, we proposed to apply chain mail jamming to the field of robotic finger design. Especially, we propose the design of the finger, the fabrication process, the method of predicting gripping force, and the grasping strategies. The experiments quantitatively verify the model of gripping force prediction. The demonstrations validate the advantages of adaptive grasp by picking a variety of items including foods, goods, and industrial components, and show the application of in-hand manipulation.

Soft optical sensing strategies are rapidly developing for soft robotic systems as a means to increase the controllability of soft compliant robots. In this paper, we present a roughness tuning strategy for the fabrication of soft optical sensors to achieve the dual functionality of shape sensing combined with contact recognition within a single multi-modal sensor. The molds used to fabricate the soft sensors are roughened via laser micromachining to achieve asymmetrical sensor responses when bent in opposite directions. We demonstrate the integration of these sensors into a fully soft robotic platform consisting of a multi-directional bending module with integrated 3D shape sensing and a gripper with tip position monitoring along with contact force recognition. We show the accuracy of our sensing strategy in validation experiments and a pick-and-place task is performed to demonstrate the robot's functionality.

Soft continuum robots have shown tremendous potential for medical and industrial applications owing to their flexibility and continuous deformability. However, their telescopic and bending capabilities and variable stiffness are still limited. This study proposes a novel origami-inspired soft continuum robot to possess large telescopic and bending capabilities while improving stiffness based on the principle of antagonistic actuation. The soft robot consists of dual origami structures. The inner forms an air chamber actuated by pneumatics, and the outer is controlled by nine tendon-driven actuators. The proposed design uses the advantages of a hybrid actuation to achieve motion and stiffness control. The performance of the soft robot is studied experimentally based on single and three robot modules. Results show that the robot has an excellent stretch ratio and a maximum bending angle of 180°. The robot can also increase stiffness to resist the bending deformation induced by self-weight and loads.

The successful integration of domestic service robots into home environments can bring significant services and convenience to the general population and possibly mitigate important societal issues, such as care provision for older adults. However, home environments are complex, dynamic and object-rich. It is, thus, very probable that service robots will encounter ambiguity while interacting with household items. To enable service robots to be more adaptive, we proposed a learning so-cial referencing computational framework and experimentally evaluated the framework on a mobile manipulator robot, Fetch, in object selection scenarios. The framework allows the robot to (1) detect and analyze the ambiguity level based on the robot's view and user's command, (2) assess the human's attention level and attract their attention, (3) disambiguate references to objects using human feedback and (4) learn novel objects after clarification from the user. System evaluation results are presented. The framework is modular and can be applied to different robotic platforms.

Soft devices employ variable stiffness to ensure safety and improve the robustness in the interaction between robots and objects. Using soft materials is one of the most popular approaches to design a variable-stiffness device, while the use of silicone sponge remains less explored in this field. Here we present a novel silicone-sponge-based variable-stiffness device (SVD). The SVD is easy-to-make and low-cost, and fabricated by an air-tight bellow enclosing a silicone sponge core. This allows easy access to the hyper-elastic response of the porous sponge whilst stiffness tuning of the device via pneumatic pressure difference. A detailed mathematical model of the SVD is proposed, by which the stiffness can be precisely controlled by the pressure difference applied. The stiffness of SVD can be tuned in the range of [\mathbf{1.55}, \mathbf{2} \mathbf{2.82}]\times \mathbf{10}^{\mathbf{3}}\ \mathbf{N}/\mathbf{m}[\mathbf{1.55}, \mathbf{2} \mathbf{2.82}]\times \mathbf{10}^{\mathbf{3}}\ \mathbf{N}/\mathbf{m}, up to 14.7 times increase. The high stiffness is easily triggered by a low pressure difference (\mathbf{\Delta} \boldsymbol{P} < \mathbf{12}\mathbf{kPa})(\mathbf{\Delta} \boldsymbol{P} < \mathbf{12}\mathbf{kPa}). The SVD is a versatile and compact module, with small axial size (10 mm height) and light weight (14.3 g), making it highly suitable for integration in a wide range of robotics and industrial applications. This, in addition to its easy-to-fabricate and low-cost features, may appeal to the robotics community at large. We further detail its working principle, fabrication processes, mathematical model and automated control methods to show its versatility.

In this work, we propose a numerical method for the solution of local generalized Nash equilibria (GNE) for the class of open-loop general-sum dynamic games for agents with nonlinear dynamics and constraints. In particular, we formulate a sequential quadratic programming (SQP) approach which requires only the solution of a single convex quadratic program at each iteration and is locally convergent. Central to the effectiveness of our approach is a non-monotonic line search method and a novel merit function for SQP step acceptance which helps to improve solver convergence beyond the local neighborhood of a GNE. We demonstrate the effectiveness of the algorithm in the context of car racing, where we see up to 32% improvement of success rate when comparing against a recent solution approach for dynamic games. We also make our code available at https://github.com/zhu-edward/DGSQP.

Closed-loop state sensitivity [1], [2] is a recently introduced notion that can be used to quantify deviations of the closed-loop trajectory of a robot/controller pair against variations of uncertain parameters in the robot model. While local optimization techniques are used in [1], [2] to generate reference trajectories minimizing a sensitivity-based cost, no global planning algorithm considering this metric to compute collision-free motions robust to parametric uncertainties has yet been proposed. The contribution of this paper is to propose a global control-aware motion planner for optimizing a state sensitivity metric and producing collision-free reference motions that are robust against parametric uncertainties for a large class of complex dynamical systems. Given the prohibitively high computational cost of directly minimizing the state sensitivity using asymptotically optimal sampling-based tree planners, the proposed RRT*-based SAMP planner uses an appropriate steering method to first compute a (near) time-optimal and kinodynamically feasible trajectory that is then locally deformed to improve robustness and decrease its sensitivity to uncertainties. The evaluation performed on planar/full-3D quadrotor UAV models shows that the SAMP method produces low sensitivity robust solutions with a much higher performance than a planner directly optimizing the sensitivity.

Autonomy, cooperation and data fusion can increase the performance of robotic networks in many underwater applications. In this paper, we describe a novel occupancy grid (OG) based perception layer, and its use for controlling a network of autonomous underwater vehicles (AUVs), sensorised with passive sonars. Data fusion between the robots' bearing-only measurements (typical of passive sonars) enables the estimate of target position. The developed OG framework exploits networking and the spatial diversity provided by the multi-robot system. The perception layer was integrated in the intelligent Cooperative Autonomous Decision Making Engine (iCADME) control architecture and validated for the first time in the Robotics Experimentation and Prototyping MUS (REPMUS) Exercise, held in Portugal in September 2021. Our robotic network participated in a technical demonstration, whose main objective was to localise a bottomed submarine which emitted a periodic help request acoustically during a simulated distress situation. We report results which are one of the first examples to demonstrate how cooperative robotics, supported by data fusion, can be effective in a passive sonar scenario. They also confirm the viability of adopting such solutions in real-world applications, characterised by poor communications and challenging environments. What was achieved at REPMUS21 clearly demonstrates how a network of cooperative robots can improve search & rescue operations of a submarine.

This paper presents a novel algorithm for robot task and motion planning (TAMP) problems by utilizing a reachability tree. While tree-based algorithms are known for their speed and simplicity in motion planning (MP), they are not well-suited for TAMP problems that involve both abstracted and geometrical state variables. To address this challenge, we propose a hierarchical sampling strategy, which first generates an abstracted task plan using Monte Carlo tree search (MCTS) and then fills in the details with a geometrically feasible motion trajectory. Moreover, we show that the performance of the proposed method can be significantly enhanced by selecting an appropriate reward for MCTS and by using a pre-generated goal state that is guaranteed to be geometrically feasible. A comparative study using TAMP benchmark problems demonstrates the effectiveness of the proposed approach.

In recent years, a deep learning framework has been widely used for object pose estimation. While quaternion is a common choice for rotation representation, it cannot represent the ambiguity of the observation. In order to handle the ambiguity, the Bingham distribution is one promising solution. However, it requires complicated calculation when yielding the negative log-likelihood (NLL) loss. An alternative easy-to-implement loss function has been proposed to avoid complex computations but has difficulty expressing symmetric distribution. In this paper, we introduce a fast-computable and easy-to-implement NLL loss function for Bingham distribution. We also create the inference network and show that our loss function can capture the symmetric property of target objects from their point clouds.

Human activity recognition has become an attractive research area with the development of on-body wearable sensing technology. With comfortable electronic-textiles, sensors can be embedded into clothing so that it is possible to record human movement outside the laboratory for long periods. However, a long-standing issue is how to deal with motion artifact introduced by movement of clothing with respect to the body. Surprisingly, recent empirical findings suggest that cloth-attached sensor can actually achieve higher accuracy of activity recognition than rigid-attached sensor, particularly when predicting from short time-windows. In this work, a probabilistic model is introduced in which this improved accuracy and resposiveness is explained by the increased statistical distance between movements recorded via fabric sensing. The predictions of the model are verified in simulated and real human motion capture experiments, where it is evident that this counterintuitive effect is closely captured.

Visual localization allows autonomous robots to relocalize when losing track of their pose by matching their current observation with past ones. However, ambiguous scenes pose a challenge for such systems, as repetitive structures can be viewed from many distinct, equally likely camera poses, which means it is not sufficient to produce a single best pose hypothesis. In this work, we propose a probabilistic framework that for a given image predicts the arbitrarily shaped posterior distribution of its camera pose. We do this via a novel formulation of camera pose regression using variational inference, which allows sampling from the predicted distribution. Our method outperforms existing methods on localization in ambiguous scenes. We open-source our approach and share our recorded data sequence at github.com/efreidun/vapor.

This paper explores the effects of a swing phase stumble recovery controller in a swing-assist prosthesis. The prosthesis detects a stumble event and employs either a lowering recovery response - wherein the user's swing is truncated, and the leg is prepared for loading- or an elevating recovery response - wherein an amplified swing flexion is employed to step over the obstacle causing the perturbation. The controller described in this paper choses which of these responses to use based on the perturbation timing within the gait cycle, where stumble events which occur prior to an estimated 35 percent of the way through swing trigger an elevating response, and later stumbles trigger a lowering response. The potential efficacy of this approach was assessed in a preliminary study with two participants with transfemoral amputation; wherein each participant's walking was perturbed in early, mid, and late swing phase when wearing both their prescribed prosthesis and the swing-assist prosthesis prototype. When wearing the swing- assist device, 0 of the 13 perturbations resulted in falls, with none of the trials being classifiable as “near falls”. Conversely, when using their prescribed device, one participant had a fall rate of 3 out of 6 perturbations, with 1 of the 3 recoveries being classifiable as a “near fall”; the second participant had a fall rate of 0 of 3 trials, with 2 of the 3 recoveries being classifiable as “near falls”. For both participants, when recovery was achieved, it was accompanied by significantly longer periods of irregularity and asymmetry in gait when using their prescribed devices, as compared to the test device. These results suggest the possibility of substantial benefit provided by a low-power, reflex- based stumble recovery feature in knee prostheses.

A plug-in weight-shifting module that can be inserted into a variety of objects is presented. The module is equipped with a movable weight inside its body. Three-dimensional weight shifts are presented by controlling one-dimensional translational and two-dimensional rotational movements. To explore the use case of this weight-shifting module, eight weight shift patterns expressing certain emotions were created through a workshop and a qualitative analysis. User tests, to which three different embodiments and scenarios were applied, examined the following three cases: the weight shift patterns were presented to the user by a) a stuffed toy-style robot that mediated human messaging, b) a cushion that made the user relax, and c) a container that enhanced the user's movie-watching experience. User interviews revealed the feasibility of the module and its weight shift patterns for the user's perception of emotions.

The high speed and low energy cost are two conflicting objectives in the motion optimization of bio-inspired underwater robots, but playing a very important role. To this end, this paper proposes an optimization strategy for swimming speed and power cost using an improved NSGA-II for a flexible robotic fish. A dynamic model involving flexible deformation is established for speed prediction with the hydrodynamic parameters identified. A back propagation (BP) neural network is applied to perform compensation of power cost prediction with the dynamic model's prediction as input. In particular, an NSGA-II-AMS method is developed to improve the efficiency of solving the two-objective optimization problem based on NSGA-II. Finally, extensive simulations and experimental results demonstrate the effectiveness of the proposed optimization strategy, which offers promising prospects for the flexible robotic fish performing aquatic tasks with different performance constraints.

As robots become more and more intelligent, the complexity of the algorithms behind them is increasing. Since these algorithms require high computation power from the onboard robot controller, the weight of the robot and energy consumption increases. A promising solution to tackle this issue is to relocate the expensive computation to the cloud. In this pioneering work, the possibility of relocating a state-of-the-art nonlinear control is investigated. To this end, the Unified Force-Impedance Controller (UFIC) is relocated to a remote location and high frequency feedback loop is established by including the remote controller in the loop. Passivity analysis is used to ensure the stability of the whole system, comprising the robot in interaction with the environment, the communication channel, as well as the remote controller. The instability associated with the communication channel is resolved by Time Domain Passivity Approach (TDPA). The performance of the proposed framework is experimentally evaluated on a robot arm in interaction with the environment. The results illustrate the stability of the system to a time-varying delay of up to 50 ± 10ms.

This paper presents a simple yet effective design of a platform to automate the task of shellfish aquaculture, specifically pearl oysters. Compared to traditional methods, our platform can eliminate the tedious task of cleaning the pearl oysters due to fouling. Inspired by the low and high tide characteristics of the intertidal zone, our platform employs an air-water displacement mechanism to periodically float pearl oysters above the water's surface, exposing fouling organisms to air and sunlight. While pearl oysters have developed the ability to stay alive during low tide, these fouling organisms cannot survive after prolonged exposure, thus preventing them from developing. Additionally, the platform provides an alternative approach to grow not only pearl oysters but also various types of shellfish, consequently benefiting the aquaculture industry. We introduce the design of the platform and provide a comprehensive analysis. We also demonstrate the practical deployment of the platform for cultivating pearl oysters.