Real-world Robot Research Articles

LoCS-Net: Localizing convolutional spiking neural network for fast visual place recognition

Visual place recognition (VPR) is the ability to recognize locations in a physical environment based only on visual inputs. It is a challenging task due to perceptual aliasing, viewpoint and appearance variations and complexity of dynamic scenes. Despite promising demonstrations, many state-of-the-art (SOTA) VPR approaches based on artificial neural networks (ANNs) suffer from computational inefficiency. However, spiking neural networks (SNNs) implemented on neuromorphic hardware are reported to have remarkable potential for more efficient solutions computationally. Still, training SOTA SNNs for VPR is often intractable on large and diverse datasets, and they typically demonstrate poor real-time operation performance. To address these shortcomings, we developed an end-to-end convolutional SNN model for VPR that leverages backpropagation for tractable training. Rate-based approximations of leaky integrate-and-fire (LIF) neurons are employed during training, which are then replaced with spiking LIF neurons during inference. The proposed method significantly outperforms existing SOTA SNNs on challenging datasets like Nordland and Oxford RobotCar, achieving 78.6% precision at 100% recall on the Nordland dataset (compared to 73.0% from the current SOTA) and 45.7% on the Oxford RobotCar dataset (compared to 20.2% from the current SOTA). Our approach offers a simpler training pipeline while yielding significant improvements in both training and inference times compared to SOTA SNNs for VPR. Hardware-in-the-loop tests using Intel's neuromorphic USB form factor, Kapoho Bay, show that our on-chip spiking models for VPR trained via the ANN-to-SNN conversion strategy continue to outperform their SNN counterparts, despite a slight but noticeable decrease in performance when transitioning from off-chip to on-chip, while offering significant energy efficiency. The results highlight the outstanding rapid prototyping and real-world deployment capabilities of this approach, showing it to be a substantial step toward more prevalent SNN-based real-world robotics solutions.

Uncanny Valley Curve or Linear Relationship? The Different Influences of Robots’ Human-Likeness on Mind Perception and Visuospatial Perspective-Taking

High but imperfect human likeness (HL) of robot appearance may elicit negative emotional responses of people, which is called the Uncanny Valley (UV) effect. Meanwhile, robots’ human-like appearances also influence people’s mind perception and behaviors toward robots, but it is not clear whether the influences also follow the UV curve. Moreover, the relationship between mind perception and behavior remains an open question. This study recruited 256 participants and employed 72 real-world robots with different levels of human-like appearance. Participants’ spontaneous visuospatial perspective-taking, along with their subjective ratings of robots’ likability and mind perception were measured. The results revealed that with the HL score of robots increasing, likability and mind perception showed UV curves, but perspective-taking exhibited an increasingly linear line. At the same time, the relationship between perspective-taking and mind perception was weak. These empirical findings provided insights into theories related to people’s mind perceptions and behaviors toward robots.

Music Mode: Transforming Robot Movement into Music Increases Likability and Perceived Intelligence

As robots enter everyday spaces like offices, the sounds they create affect how they are perceived. We present “Music Mode,” a novel mapping between a robot’s joint motions and sounds, programmed by artists and engineers to make the robot generate music as it moves. Two experiments were designed to characterize the effect of this musical augmentation on human users. In the first experiment, a robot performed three tasks while playing three different sound mappings. Results showed that participants observing the robot perceived it as more safe, animate, intelligent, anthropomorphic, and likable when playing the Music Mode Orchestra software. To test whether the results of the first experiment were due to the Music Mode algorithm, rather than music alone, we conducted a second experiment. Here, the robot performed the same three tasks, while a participant observed via video, but the Orchestra music was either linked to its movement or random. Participants rated the robots as more intelligent when the music was linked to the movement. Robots using Music Mode logged approximately 200 hours of operation while navigating, wiping tables, and sorting trash, and bystander comments made during this operating time served as an embedded case study. This article has both designerly contributions and engineering contributions. The contributions are as follows: (1) an interdisciplinary choreographic, musical, and coding design process to develop a real-world robot sound feature, (2) a technical implementation for movement-based sound generation, and (3) two experiments and an embedded case study of robots running this feature during daily work activities that resulted in increased likeability and perceived intelligence of the robot.

Reinforcement Learning with Decoupled State Representation for Robot Manipulations

Deep reinforcement learning has significantly advanced robot manipulations by providing an alternative solution for designing control strategies using raw images as direct inputs. While images offer additional environmental information, the end-to-end policy training manner (from image to action) requires simultaneous representation and task learning by the agent. This often necessitates a substantial number of interaction samples to achieve satisfactory policy performance. Previous works has attempted to address this challenge by learning a visual representation model that encodes the entire image into a low-dimensional vector before the policy training. However, since this vector contains both robot and object information, it inevitably introduces coupling within the state, which can mislead the policy training process. In this study, a novel method called Reinforcement Learning with Decoupled State Representation is proposed to effectively decouple robot and object information within the state representation. Experimental results demonstrate that the proposed method exhibits faster learning speed and achieves superior performance compared to previous methods across various robot manipulation tasks. Moreover, with only 3096 offline images, the proposed method successfully applies to real-world robot pushing tasks, which demonstrates its high practicability.

Special issue on real-world robot applications of the foundation models

Special issue on real-world robot applications of the foundation models

Minimum-energy switching geometric filter on lie groups for differential-drive wheeled mobile robots

Accurate state estimation plays a critical role in various applications, such as tracking, regulation, and fault detection in robotic and mechanical systems. Typically, the Kalman–Bucy filter is used as a linear state observer for this purpose. However, real-world robots often exhibit complex behavior, characterized by a combination of dynamics, making it essential to employ hybrid filters. In this context, the Switching Kalman filter stands out as a well-established solution. In this article we aim to generalize the Brownian-Markov Stochastic Model, a hybrid dynamic model for differential-drive wheeled mobile robots, to the case of a mobile robot whose center of mass is not aligned to the wheels axle middle point, and to design a geometric hybrid state estimator by exploiting the Lie groups theory. The Brownian-Markov Stochastic Model features two modes: “grip” and “slip”. These modes correspond to ideal grip and lateral slippage, with transitions governed by a state-dependent Markov chain. To validate the proposed switching filter, we conduct MATLAB® simulations of the robot’s motion in a scenario prone to lateral grip loss, comparing the state estimates produced by the switching geometric filter with those obtained using the switching Kalman filter.

Using Reinforcement Learning to Develop a Novel Gait for a Bio-Robotic California Sea Lion.

While researchers have made notable progress in bio-inspired swimming robot development, a persistent challenge lies in creating propulsive gaits tailored to these robotic systems. The California sea lion achieves its robust swimming abilities through a careful coordination of foreflippers and body segments. In this paper, reinforcement learning (RL) was used to develop a novel sea lion foreflipper gait for a bio-robotic swimmer using a numerically modelled computational representation of the robot. This model integration enabled reinforcement learning to develop desired swimming gaits in the challenging underwater domain. The novel RL gait outperformed the characteristic sea lion foreflipper gait in the simulated underwater domain. When applied to the real-world robot, the RL constructed novel gait performed as well as or better than the characteristic sea lion gait in many factors. This work shows the potential for using complimentary bio-robotic and numerical models with reinforcement learning to enable the development of effective gaits and maneuvers for underwater swimming vehicles.

Learning periodic skills for robotic manipulation: Insights on orientation and impedance

Many daily tasks exhibit a periodic nature, necessitating that robots possess the ability to execute them either alone or in collaboration with humans. A widely used approach to encode and learn such periodic patterns from human demonstrations is through periodic Dynamic Movement Primitives (DMPs). Periodic DMPs encode cyclic data independently across multiple dimensions of multi-degree of freedom systems. This method is effective for simple data, like Cartesian or joint position trajectories. However, it cannot account for various geometric constraints imposed by more complex data, such as orientation and stiffness. To bridge this gap, we propose a novel periodic DMP formulation that enables the encoding of periodic orientation trajectories and varying stiffness matrices while considering their geometric constraints. Our geometry-aware approach exploits the properties of the Riemannian manifold and Lie group to directly encode such periodic data while respecting its inherent geometric constraints. We initially employed simulation to validate the technical aspects of the proposed method thoroughly. Subsequently, we conducted experiments with two different real-world robots performing daily tasks involving periodic changes in orientation and/or stiffness, i.e., operating a drilling machine using a rotary handle and facilitating collaborative human–robot sawing.

Learning accurate and efficient three-finger grasp generation in clutters with an auto-annotated large-scale dataset

With the development of intelligent manufacturing and robotic technologies, the capability of grasping unknown objects in unstructured environments is becoming more prominent for robots with extensive applications. However, current robotic three-finger grasping studies only focus on grasp generation for single objects or scattered scenes, and suffer from high time expenditure to label grasp ground truth, making them incapable of predicting grasp poses for cluttered objects or generating large-scale datasets. To address such limitations, we first introduce a novel three-finger grasp representation with fewer prediction dimensions, which balances the training difficulty and representation accuracy to obtain efficient grasp performance. Based on this representation, we develop an auto-annotation pipeline and contribute a large-scale three-finger grasp dataset (TF-Grasp Dataset). Our dataset contains 222,720 RGB-D images with over 2 billion grasp annotations in cluttered scenes. In addition, we also propose a three-finger grasp pose detection network (TF-GPD), which detects globally while fine-tuning locally to predict high-quality collision-free grasps from a single-view point cloud. In sum, our work addresses the issue of high-quality collision-free three-finger grasp generation in cluttered scenes based on the proposed pipeline. Extensive comparative experiments show that our proposed methodology outperforms previous methods and improves the grasp quality and efficiency in clutters. The superior results in real-world robot grasping experiments not only prove the reliability of our grasp model but also pave the way for practical applications of three-finger grasping. Our dataset and source code will be released.

Enhancing Autonomous Driving Robot Systems with Edge Computing and LDM Platforms

The efficient operation and interaction of autonomous robots play crucial roles in various fields, e.g., security, environmental monitoring, and disaster response. For these purposes, processing large volumes of sensor data and sharing data between robots is essential; however, processing such large data in an on-device environment for robots results in substantial computational resource demands, causing high battery consumption and heat issues. Thus, this study addresses challenges related to processing large volumes of sensor data and the lack of dynamic object information sharing among autonomous robots and other mobility systems. To this end, we propose an Edge-Driving Robotics Platform (EDRP) and a Local Dynamic Map Platform (LDMP) based on 5G mobile edge computing and Kubernetes. The proposed EDRP implements the functions of autonomous robots based on a microservice architecture and offloads these functions to an edge cloud computing environment. The LDMP collects and shares information about dynamic objects based on the ETSI TR 103 324 standard, ensuring cooperation among robots in a cluster and compatibility with various Cooperative-Intelligent Transport System (C-ITS) components. The feasibility of operating a large-scale autonomous robot offloading system was verified in experimental scenarios involving robot autonomy, dynamic object collection, and distribution by integrating real-world robots with an edge computing–based offloading platform. Experimental results confirmed the potential of dynamic object collection and dynamic object information sharing with C-ITS environment components based on LDMP.

Improved Hybrid Model for Obstacle Detection and Avoidance in Robot Operating System Framework (Rapidly Exploring Random Tree and Dynamic Windows Approach).

The integration of machine learning and robotics brings promising potential to tackle the application challenges of mobile robot navigation in industries. The real-world environment is highly dynamic and unpredictable, with increasing necessities for efficiency and safety. This demands a multi-faceted approach that combines advanced sensing, robust obstacle detection, and avoidance mechanisms for an effective robot navigation experience. While hybrid methods with default robot operating system (ROS) navigation stack have demonstrated significant results, their performance in real time and highly dynamic environments remains a challenge. These environments are characterized by continuously changing conditions, which can impact the precision of obstacle detection systems and efficient avoidance control decision-making processes. In response to these challenges, this paper presents a novel solution that combines a rapidly exploring random tree (RRT)-integrated ROS navigation stack and a pre-trained YOLOv7 object detection model to enhance the capability of the developed work on the NAV-YOLO system. The proposed approach leveraged the high accuracy of YOLOv7 obstacle detection and the efficient path-planning capabilities of RRT and dynamic windows approach (DWA) to improve the navigation performance of mobile robots in real-world complex and dynamically changing settings. Extensive simulation and real-world robot platform experiments were conducted to evaluate the efficiency of the proposed solution. The result demonstrated a high-level obstacle avoidance capability, ensuring the safety and efficiency of mobile robot navigation operations in aviation environments.

A fictional history of robotics features forgotten real-world robots.

The science-fiction movie The Creator uses six real-world robots from the 1950s and 1960s to show progress in AI.

Human skill knowledge guided global trajectory policy reinforcement learning method.

Traditional trajectory learning methods based on Imitation Learning (IL) only learn the existing trajectory knowledge from human demonstration. In this way, it can not adapt the trajectory knowledge to the task environment by interacting with the environment and fine-tuning the policy. To address this problem, a global trajectory learning method which combinines IL with Reinforcement Learning (RL) to adapt the knowledge policy to the environment is proposed. In this paper, IL is proposed to acquire basic trajectory skills, and then learns the agent will explore and exploit more policy which is applicable to the current environment by RL. The basic trajectory skills include the knowledge policy and the time stage information in the whole task space to help learn the time series of the trajectory, and are used to guide the subsequent RL process. Notably, neural networks are not used to model the action policy and the Q value of RL during the RL process. Instead, they are sampled and updated in the whole task space and then transferred to the networks after the RL process through Behavior Cloning (BC) to get continuous and smooth global trajectory policy. The feasibility and the effectiveness of the method was validated in a custom Gym environment of a flower drawing task. And then, we executed the learned policy in the real-world robot drawing experiment.

Task-Oriented Self-Imitation Learning for Robotic Autonomous Skill Acquisition

The inferior sample efficiency of reinforcement learning (RL) and the requirement for high-quality demonstrations in imitation learning (IL) will hinder their application in real-world robots. To address this challenge, a novel self-evolution framework, named task-oriented self-imitation learning (TOSIL), is proposed. To circumvent external demonstrations, the top-K self-generated trajectories are chosen as expert data from both per-episode exploration and long-term return perspectives. Each transition is assigned a guide reward, which is formulated by these trajectories. The guide rewards update as the agent evolves, encouraging good exploration behaviors. This methodology guarantees that the agent explores in the direction relevant to the task, improving sample efficiency and asymptotic performance. The experimental results on locomotion and manipulation tasks indicate that the proposed framework outperforms other state-of-the-art RL methods. Furthermore, the integration of suboptimal trajectories has the potential to improve the sample efficiency while maintaining performance. This is a significant advancement in autonomous skill acquisition for robots.

Self-organized swarm robot for multi-target trapping based on self-regulated density interaction

In swarm robotics, multi-target trapping usually relies on global information or explicit communication, posing a challenge for robots to autonomously self-organize and trap multiple targets with only local perceptual data. We present a self-regulated density-based approach for self-organized multi-target trapping. This method employs density-based negative feedback control and density-distance weighted target selection. Our approach leverages distributed perception, where robots perceive nearby peers within their sensory range, eliminating the need for explicit information exchange and enabling autonomous trapping. During task execution, negative feedback control adjusts swarm density, ensuring it reaches an optimal level. Building on this, individual robots use density fields and relative target positions to select suitable targets, achieving self-regulated dispersed selection and multi-target trapping. We validate our approach through numerical simulations and real-world robot experiments. Results demonstrate stable self-organization, efficient self-regulated dispersed target selection, and successful target trapping, even with a limited number of trapping robots in low-redundancy scenarios.

Self-reflective terrain-aware robot adaptation for consistent off-road ground navigation

Ground robots require the crucial capability of traversing unstructured and unprepared terrains and avoiding obstacles to complete tasks in real-world robotics applications such as disaster response. When a robot operates in off-road field environments such as forests, the robot’s actual behaviors often do not match its expected or planned behaviors, due to changes in the characteristics of terrains and the robot itself. Therefore, the capability of robot adaptation for consistent behavior generation is essential for maneuverability on unstructured off-road terrains. In order to address the challenge, we propose a novel method of self-reflective terrain-aware adaptation for ground robots to generate consistent controls to navigate over unstructured off-road terrains, which enables robots to more accurately execute the expected behaviors through robot self-reflection while adapting to varying unstructured terrains. To evaluate our method’s performance, we conduct extensive experiments using real ground robots with various functionality changes over diverse unstructured off-road terrains. The comprehensive experimental results have shown that our self-reflective terrain-aware adaptation method enables ground robots to generate consistent navigational behaviors and outperforms the compared previous and baseline techniques.

Towards Autonomous, Optimal Water Sampling with Aerial and Surface Vehicles for Rapid Water Quality Assessment

Highlights A practical workflow for optimizing sampling tours for a team of surface and aerial vehicles was developed. Proposed workflow considers unique sensing capabilities of surface vehicles when assigning sampling locations. Likely optimal tours can be found in less than 30 s for practical water quality sampling requirements. Abstract. Most current marine aquaculture operations are located in coastal estuarine areas within one mile of the shoreline, and water quality in these production areas can quickly become unfavorable due to hydrodynamic processes and excessive runoff. The deployment of autonomous, robotic systems can improve the speed and spatiotemporal resolution of water sampling and sensing in mariculture production areas to assess water quality in the context of food safety. Specifically, teams of both aerial and surface vehicles can be deployed simultaneously to capitalize on the benefits of each system; however, a method to optimally design a feasible sampling tour for each robot is needed to maximize sample capacity and ensure efficient water sampling missions. This research brief presents the problem formulation and a solution method to determine optimal tours for a team of aquatic surface and aerial vehicles while considering different vehicle sampling capacities and endurance constraints. This method was implemented to design sampling missions of 15, 20, and 30 samples in both a 0.25 km2 and 3.9 km2 site, using sampling capacity and endurance constraints corresponding to real-world robots used for water sampling in mariculture environments. Results indicate that this optimization problem can be solved in near-real time in the field and yields feasible sampling tours for surface and aerial vehicles under different constraints. This work is a practical step towards developing teams of collaborative robots to persistently monitor adverse mariculture growing conditions so producers can implement data-driven, timely management strategies. Keywords: Mariculture, Robotics, Traveling salesperson problem, Vehicle routing.

Real-World Robot Control Based on Contrastive Deep Active Inference With Demonstrations

Real-World Robot Control Based on Contrastive Deep Active Inference With Demonstrations

Improving Offline Reinforcement Learning With In-Sample Advantage Regularization for Robot Manipulation.

Offline reinforcement learning (RL) aims to learn the possible policy from a fixed dataset without real-time interactions with the environment. By avoiding the risky exploration of the robot, this approach is expected to significantly improve the robot's learning efficiency and safety. However, due to errors in value estimation from out-of-distribution actions, most offline RL algorithms constrain or regularize the policy to the actions contained within the dataset. The cost of such methods is the introduction of new hyperparameters and additional complexity. In this article, we aim to adapt offline RL to robotic manipulation with minimal changes and to avoid evaluating out-of-distribution actions as much as possible. Therefore, we improve offline RL with in-sample advantage regularization (ISAR). To mitigate the impact of unseen actions, the ISAR learns the state-value function only with the dataset sample to regress the optimal action-value function. Our method calculates the advantage function of action-state pairs based on in-sample value estimation and adds a behavior cloning (BC) regularization term in the policy update. This improves sample efficiency with minimal changes, resulting in a simple and easy-to-implement method. The experiments of the D4RL robot benchmark and multigoal sparse rewards robotic tasks show that the ISAR achieves excellent performance comparable to current state-of-the-art algorithms without the need for complex parameter tuning and too much training time. In addition, we demonstrate the effectiveness of our method on a real-world robot platform.

Privacy and utility perceptions of social robots in healthcare

Many Human-Robot Interaction (HRI) researchers are exploring the use of healthcare robots. Due to the sensitive nature of care, privacy concerns play a significant role in determining robot utility and adoption. While HRI research has explored some dimensions of privacy for robots in general, to our knowledge, no prior work has empirically studied how human-like robot design affects people's privacy and utility perceptions of robots across different healthcare contexts and tasks. We conducted a 3 × 3 × 3 study (n = 239) to understand these relationships, varying robot Human Likeness (HL) (low, medium, and high) and scenario/task type (hospital waiting room/robot check-in support, hospital patient room/robot mobility support, home care/robot neurorehabilitation support) via a mixed between-within subjects design. To our knowledge, this is one of the first studies that operationalizes complex constructs of privacy, healthcare, and HL across multiple realistic healthcare contexts, with a high degree of cognitive fidelity. Our results suggest the tasks and contexts in which privacy is considered in healthcare contexts with robots is more impactful than other factors like robot HL appearance. In particular, some settings include more complex tradeoffs between privacy and utility for robots than others. For example, HRI researchers and practitioners who want to build healthcare robots intended for the home may encounter the greatest challenges for balancing privacy risks. Finally, for the community, we demonstrate that design fiction animations can be a useful way to facilitate cognitive fidelity for supporting studies in HRI and serving as a bridge between narrative methods and the use of real-world robots.