Physical Robot Research Articles

Learning of Long-Horizon Sparse-Reward Robotic Manipulator Tasks With Base Controllers.

Deep reinforcement learning (DRL) enables robots to perform some intelligent tasks end-to-end. However, there are still many challenges for long-horizon sparse-reward robotic manipulator tasks. On the one hand, a sparse-reward setting causes exploration inefficient. On the other hand, exploration using physical robots is of high cost and unsafe. In this article, we propose a method of learning long-horizon sparse-reward tasks utilizing one or more existing traditional controllers named base controllers in this article. Built upon deep deterministic policy gradients (DDPGs), our algorithm incorporates the existing base controllers into stages of exploration, value learning, and policy update. Furthermore, we present a straightforward way of synthesizing different base controllers to integrate their strengths. Through experiments ranging from stacking blocks to cups, it is demonstrated that the learned state-based or image-based policies steadily outperform base controllers. Compared to previous works of learning from demonstrations, our method improves sample efficiency by orders of magnitude and improves performance. Overall, our method bears the potential of leveraging existing industrial robot manipulation systems to build more flexible and intelligent controllers.

User Study to Validate the Performance of an Offline Robot Programming Method That Enables Robot-Independent Kinesthetic Instruction through the Use of Augmented Reality and Motion Capturing

The paper presents a novel offline programming (OLP) method based on programming by demonstration (PbD), which has been validated through user study. PbD is a programming method that involves physical interaction with robots, and kinesthetic teaching (KT) is a commonly used online programming method in industry. However, online programming methods consume significant robot resources, limiting the speed advantages of PbD and emphasizing the need for an offline approach. The method presented here, based on KT, uses a virtual representation instead of a physical robot, allowing independent programming regardless of the working environment. It employs haptic input devices to teach a simulated robot in augmented reality and uses automatic path planning. A benchmarking test was conducted to standardize equipment, procedures, and evaluation techniques to compare different PbD approaches. The results indicate a 47% decrease in programming time when compared to traditional KT methods in established industrial systems. Although the accuracy is not yet at the level of industrial systems, users have shown rapid improvement, confirming the learnability of the system. User feedback on the perceived workload and the ease of use was positive. In conclusion, this method has potential for industrial use due to its learnability, reduction in robot downtime, and applicability across different robot sizes and types.

Exploring Maze Navigation: A Comparative Study of DFS, BFS, and A* Search Algorithms.

A comparative study was conducted using Python and Pmaze to evaluate the performance of DFS, BFS, and A* search algorithms in path planning for a maze. The main objective of the project was to compare the efficiency of these algorithms in terms of path cost and algorithmic complexity. No physical robot was used in this study as it focused solely on the analysis of search algorithms. This study contributes to the advancement of knowledge in the field of path planning by exploring different approaches for mazes using Python and Pmaze.

Digital twinning of surveillance robot

Surveillance robots provide troops with real-time information about their surroundings, including enemy positions, terrain, and potential threats. This information is invaluable for making informed decisions and ensuring the safety of military personnel. Geo-fenced robots are robots equipped with technology that restricts their movements within predefined geographic boundaries. These boundaries are typically established using GPS or other location-based technologies. In applications like the military, geo-fencing can be used to establish secure zones. Geo-fencing helps prevent robots from entering hazardous areas, reducing the risk of damage to the robot and potential harm to friendly forces or civilians This project aims at Digital twinning of robots by creating a virtual replica or model of a physical robot in a digital environment. Digital twinning allows engineers and designers to simulate and test the robot’s behavior, performance, and capabilities in a virtual environment before building the physical robot. This can lead to increased efficiency, reduced downtime, and cost savings. In this project, LiDAR (Light Detection and Ranging) sensor data is integrated with a digital twinned robot to create a virtual reproduction of the robot and its surroundings. LiDAR is a remote sensing technology that maps the robot’s surroundings in fine detail using 3D point clouds by measuring distances using laser pulses. Here we make use of RPLIDAR A1 M8 and acquire data from it using ROS with the help of a RaspberryPi4B Controller. Simulink is used to create a 3D model of the robot’s environment and the robot itself. Reinforcement learning and pure pursuit algorithms are used for developing them. This project discusses the need for geofenced autonomous robots and emphasizes the security and reliability it brings to military applications.

Fortune-Telling Finches: Linear Functions as Predictors

This lesson gives students experience with block coding Finch robots to move across plexiglass surfaces. It builds student computational thinking skills, mathematical self-efficacy, and interest in technology careers (McLurkin, et al., 2013; Martin, 2019; Kazi, 2023). Finch robots are small, physical robots that can be coded to move in a space. In this 8th grade math unit, conducted in the library, students identify relationships between variables while using block coding to make predictions and test hypotheses. This lesson extends students' knowledge of linear functions through observing patterns and using the guess and check strategy to complete a variety of challenges.

Open continuum robotics-one actuation module to create them all.

Experiments on physical continuum robot are the gold standard for evaluations. Currently, as no commercial continuum robot platform is available, a large variety of early-stage prototypes exists. These prototypes are developed by individual research groups and are often used for a single publication. Thus, a significant amount of time is devoted to creating proprietary hardware and software hindering the development of a common platform, and shifting away scarce time and efforts from the main research challenges. We address this problem by proposing an open-source actuation module, which can be used to build different types of continuum robots. It consists of a high-torque brushless electric motor, a high resolution optical encoder, and a low-gear-ratio transmission. For this article, we create three different types of continuum robots. In addition, we illustrate, for the first time, that continuum robots built with our actuation module can proprioceptively detect external forces. Consequently, our approach opens untapped and under-investigated research directions related to the dynamics and advanced control of continuum robots, where sensing the generalized flow and effort is mandatory. Besides that, we democratize continuum robots research by providing open-source software and hardware with our initiative called the Open Continuum Robotics Project, to increase the accessibility and reproducibility of advanced methods.

Virtual Shake Robot: Simulating Dynamics of Precariously Balanced Rocks for Overturning and Large-displacement Processes

Understanding the dynamics of precariously balanced rocks (PBRs) is important for seismic hazard analysis and rockfall prediction. Utilizing a physics engine and robotic tools, we develop a virtual shake robot (VSR) to simulate the dynamics of PBRs during overturning and large-displacement processes. We present the background of physics engines and technical details of the VSR, including software architecture, mechanical structure, control system, and implementation procedures. Validation experiments show the median fragility contour from VSR simulation is within the 95% prediction intervals from previous physical experiments, when PGV/PGA is greater than 0.08 s. Using a physical mini shake robot, we validate the qualitative consistency of fragility anisotropy between the VSR and physical experiments. By overturning cuboids on flat terrain, the VSR reveals the relationship between fragility and geometric dimensions (e.g., aspect and scaling ratios). The ground motion orientation and lateral pedestal support affect PBR fragility. Large-displacement experiments estimate rock trajectories for different ground motions, which is useful for understanding the fate of toppled PBRs. Ground motions positively correlate with large displacement statistics such as mean trajectory length, mean largest velocity, and mean terminal distance. The overturning and large displacement processes of PBRs provide complementary methods of ground motion estimation.

Robust passivity-based nonlinear controller design for bilateral teleoperation system under variable time delay and variable load disturbance

Abstract This research focuses on implementing a robust passivity-based nonlinear control method for bilateral/teleoperation systems. The key challenge is addressing communication pathways between the master and slave, control delays, and load disturbances, which can lead to instability and reduced transparency. To tackle these issues, the proposed controller incorporates a second-order super-twisting sliding-mode observer to counteract communication and control delays. A sliding mode assist disturbance observer compensates for load torque variations. The approach aims to ensure stability and transparency by handling time-varying delays. The system model comprises two interconnected direct-drive motors, simulating robotic configurations without a physical robot. The nonlinear controller framework simplifies the complex bilateral control problem, significantly improving stability and transparency performance. Computer simulations with step and sinusoidal inputs demonstrate the effectiveness of the approach, providing a satisfactory level of accuracy and transparency between estimated and actual slave positions, even with varying delays and load variations. The research contributes to control engineering by offering a robust method to enhance bilateral system performance, ensuring stable and transparent communication between the master and slave, particularly suitable for real-time internet-based bilateral control systems.

Application Scenarios and Forms of Artificial Intelligence in Physical Education

The application of artificial intelligence in physical education is analyzed. The manifestations of artificial intelligence in physical education are as follows: accurate diagnosis, process monitoring, personalized service and intelligent decision-making; The application forms are as follows: intelligent tutor system, automatic evaluation system and physical education robot. The prospect of artificial intelligence in physical education is put forward: facing modernization and building a learning society; Facing the world, changing the teaching concept; Facing the future, innovating physical education teaching methods.

The Application of CoppeliaSim EDU in Robot Course Teaching

CoppeliaSim EDU, as a powerful tool for robot teaching, is changing the traditional teaching mode. This software provides a comprehensive robot simulation environment where users can explore different scenarios, models, and programming interfaces to gain a deeper understanding of robot technology. Its advantage lies in reducing the cost and potential security risks of physical robots, making robot education more widespread and secure. In actual teaching, teachers can use CoppeliaSim EDU for course design, guide students to carry out robot programming and simulation experiments, and improve teaching effectiveness. Through case analysis, it can be found that using CoppeliaSim EDU can better stimulate students' interest in learning, improve their practical ability and innovative thinking. In the future, with the continuous development of robotics technology, the application prospects of CoppeliaSim EDU will be even broader, bringing more possibilities for robot education.

Closing the Loop on Concentric Tube Robot Design: A Case Study on Micro-Laryngeal Surgery.

Concentric tube robots (CTRs) are well-suited to address the unique challenges of minimally invasive surgical procedures due to their small size and ability to navigate highly constrained environments. However, uncertainties in the manufacturing process can lead to challenges in the transition from simulated designs to physical robots. In this work, we propose an end-to-end design workflow for CTRs that considers the oftenoverlooked impact of manufacturing uncertainty, focusing on two primary sources - tube curvature and diameter. This comprehensive approach incorporates a two-step design optimization and an uncertainty-based selection of manufacturing tolerances. Simulation results highlight the substantial influence of manufacturing uncertainties, particularly tube curvature, on the physical robot's performance. By integrating these uncertainties into the design process, we can effectively bridge the gap between simulation and real-world performance. Two hardware experiments validate the proposed CTR design workflow. The first experiment confirms that the performance of the physical robot lies within the simulated probability distribution from the optimization, while the second experiment demonstrates the feasibility of the overall system for use in micro-laryngeal surgical tasks. This work not only contributes to a more comprehensive understanding of CTR design by addressing manufacturing uncertainties, but also creates a new framework for robust design, as illustrated in the context of microlaryngeal surgery.

Open Remote Web Lab for Learning Robotics and ROS With Physical and Simulated Robots in an Authentic Developer Environment

Open Remote Web Lab for Learning Robotics and ROS With Physical and Simulated Robots in an Authentic Developer Environment

A Real-Time Grasping Detection Network Architecture for Various Grasping Scenarios.

In the field of robot grasping detection, due to uncertain factors such as different shapes, distinct colors, diverse materials, and various poses, robot grasping has become very challenging. This article introduces a integrated robotic system designed to address the challenge of grasping numerous unknown objects within a scene from a set of α -channel images. We propose a lightweight and object-independent pixel-level generative adaptive residual depthwise separable convolutional neural network (GARDSCN) with an inference speed of around 28 ms, which can be applied to real-time grasping detection. It can effectively deal with the grasping detection of unknown objects with different shapes and poses in various scenes and overcome the limitations of current robot grasping technology. The proposed network achieves 98.88 % grasp detection accuracy on the Cornell dataset and 95.23 % on the Jacquard dataset. To further verify the validity, the grasping experiment is conducted on a physical robot Kinova Gen2, and the grasp success rate is 96.67 % in the single-object scene and 94.10 % in the multiobject cluttered scene.

A matter of consequences

Abstract On reviewing the literature regarding acceptance and trust in human-robot interaction (HRI), there are a number of open questions that needed to be addressed in order to establish effective collaborations between humans and robots in real-world applications. In particular, we identified four principal open areas that should be investigated to create guidelines for the successful deployment of robots in the wild. These areas are focused on: (1) the robot’s abilities and limitations; in particular when it makes errors with different severity of consequences, (2) individual differences, (3) the dynamics of human-robot trust, and (4) the interaction between humans and robots over time. In this paper, we present two very similar studies, one with a virtual robot with human-like abilities, and one with a Care-O-bot 4 robot. In the first study, we create an immersive narrative using an interactive storyboard to collect responses of 154 participants. In the second study, 6 participants had repeated interactions over three weeks with a physical robot. We summarise and discuss the findings of our investigations of the effects of robots’ errors on people’s trust in robots for designing mechanisms that allow robots to recover from a breach of trust. In particular, we observed that robots’ errors had greater impact on people’s trust in the robot when the errors were made at the beginning of the interaction and had severe consequences. Our results also provided insights on how these errors vary according to the individuals’ personalities, expectations and previous experiences.

Differential Outcomes Training of Visuospatial Memory: A Gamified Approach Using a Socially Assistive Robot

This study represents the first stage of evaluating whether cognitive training interventions may be facilitated by the presence of a socially assistive robot (SAR) and gamification. Our experimental setup involves using a SAR providing feedback to a gamified visuospatial working memory task, administered according to a differential outcomes training (DOT) protocol. The study’s main objective was to investigate whether performance and attitude towards the task would be affected by different robotic setups (none, simulated or physical) and in relation to different challenge levels. We measured performance accuracy on the gamified visuospatial memory task and self-reported affective ratings, which are relevant for assessing attitude towards the task and providing indicators to the potential for using a SAR for a longer-term cognitive intervention. Additionally, we conducted exploratory analyses of eye movement strategies for memory encoding during the task. The results demonstrated a significant differential outcomes effect (DOE) on memory performance accuracy, regardless of Robot type and Challenge level, providing evidence that a DOE can still be obtained when a SAR interacts with participants. Moreover, the results from the affective ratings revealed that participants accompanied by the physical robot reported lower levels of stress and increased levels of control. Our results demonstrate, for the first time, a DOE using a SAR in a gamified context. This result, coupled with positive subjective reporting of the human–robot interactive experience of participants, demonstrates the potential for using a SAR to: (i) promote positive attitudes for a DOT-based cognitive intervention, without (ii) negatively affecting task performance.

Learning and Reusing Quadruped Robot Movement Skills from Biological Dogs for Higher-Level Tasks.

In the field of quadruped robots, the most classic motion control algorithm is based on model prediction control (MPC). However, this method poses challenges as it necessitates the precise construction of the robot's dynamics model, making it difficult to achieve agile movements similar to those of a biological dog. Due to these limitations, researchers are increasingly turning to model-free learning methods, which significantly reduce the difficulty of modeling and engineering debugging and simultaneously reduce real-time optimization computational burden. Inspired by the growth process of humans and animals, from learning to walk to fluent movements, this article proposes a hierarchical reinforcement learning framework for the motion controller to learn some higher-level tasks. First, some basic motion skills can be learned from motion data captured from a biological dog. Then, with these learned basic motion skills as a foundation, the quadruped robot can focus on learning higher-level tasks without starting from low-level kinematics, which saves redundant training time. By utilizing domain randomization techniques during the training process, the trained policy function can be directly transferred to a physical robot without modification, and the resulting controller can perform more biomimetic movements. By implementing the method proposed in this article, the agility and adaptability of the quadruped robot can be maximally utilized to achieve efficient operations in complex terrains.

Marine algae inspired dispersion of swarm robots with binary sensory information

The dynamics of swarm robotic systems are complex and often nonlinear. One key issue is to design the controllers of a large number of simple, low-cost robots so that emergence can be observed. This paper presents a sensor and computation-friendly controller for swarm robotic systems inspired by the mechanisms observed in algae. The aim is to achieve uniform dispersion of robots by mimicking the circular movement observed in marine algae systems. The proposed controller utilizes binary sensory information (i.e., see or not see) to guide the robots’ motion. By moving circularly and switching the radii based on the perception of other robots in their line of sight, the robots imitate the repulsion behavior observed in algae. The controller relies solely on binary-state sensory input, eliminating the need for additional memory or communication. Up to 1024 simulated robots are used to validate the effectiveness of the dispersion controller, while experiments with 30 physical robots demonstrate the feasibility of the proposed approach.

Моделювання автономного проходження лабіринту з використанням алгоритму NEAT

The article deals with the problem of finding a solution for the navigational task of navigating a maze by an autonomous agent controlled by an artificial neural network (ANN). A solution to this problem was proposed by training the controlling ANN using the method of neuroevolution of augmenting topologies (NEAT). A description of the mathematical apparatus for determining the goal-oriented objective function to measure fitness of the decision-making agent, suitable for optimizing the training of ANN in the process of neuroevolution, was given. Based on the invented objective function, a software was developed to control the neuroevolutionary process using the Python programming language. A system for simulating the behavior of an autonomous robot that can navigate through a maze using input signals from various types of sensors has been created. The simulation system allows to imitate the behavior of a physical robot in a large number of experiments in a short time and with minimal expenses. The experiments performed using the created simulation system to find the optimal values of hyperparameters, which can be used for successful training of the controlling ANN by the method of neuroevolution, are presented. Additionally, the implemented new methods of visualizing the training process are described. These methods significantly simplify the search for optimal hyperparameters of the NEAT algorithm, due to the visual demonstration of the effect of changing one or another parameter on the training process.

Environmental force sensing helps robots traverse cluttered large obstacles

Robots can traverse sparse obstacles by sensing environmental geometry and avoiding contact with obstacles. However, for search and rescue in rubble, environmental monitoring through dense vegetation, and planetary exploration over Martian and lunar rocks, robots must traverse cluttered obstacles as large as themselves by physically interacting with them. Previous work discovered that the forest floor-dwelling discoid cockroach and a sensor-less minimalistic robot can traverse cluttered grass-like beam obstacles of various stiffness by transitioning across different locomotor modes. Yet the animal was better at traversal than the sensor-less robot, likely by sensing forces during obstacle interaction to control its locomotor transitions. Inspired by this, here we demonstrated in simulation that environmental force sensing helps robots traverse cluttered large obstacles. First, we developed a multi-body dynamics simulation and a physics model of the minimalistic robot interacting with beams to estimate beam stiffness from the sensed contact forces. Then, we developed a force feedback strategy for the robot to use the sensed beam stiffness to choose the locomotor mode with a lower mechanical energy cost. With feedforward pushing, the robot was stuck in front of stiff beams if it has a limited force capacity; without force limit, it traversed but suffered a high energy cost. Using obstacle avoidance, the robot traversed beams by avoiding beam contact regardless of beam stiffness, resulting in a high energy cost for flimsy beams. With force feedback, the robot determined beam stiffness, then traversed flimsy beams by pushing them over and stiff beams by rolling through the gap between them with a low energy cost. Stiffness estimation based on force sensing was accurate across varied body oscillation amplitude and frequency and position sensing uncertainty. Mechanical energy cost of traversal increased with sensorimotor delay. Future work should demonstrate cluttered large obstacle traversal using force feedback in a physical robot.

Multi-modal interaction with transformers: bridging robots and human with natural language

AbstractThe language-guided visual robotic grasping task focuses on enabling robots to grasp objects based on human language instructions. However, real-world human-robot collaboration tasks often involve situations with ambiguous language instructions and complex scenarios. These challenges arise in the understanding of linguistic queries, discrimination of key concepts in visual and language information, and generation of executable grasping configurations for the robot’s end-effector. To overcome these challenges, we propose a novel multi-modal transformer-based framework in this study, which assists robots in localizing spatial interactions of objects using text queries and visual sensing. This framework facilitates object grasping in accordance with human instructions. Our developed framework consists of two main components. First, a visual-linguistic transformer encoder is employed to model multi-modal interactions for objects referred to in the text. Second, the framework performs joint spatial localization and grasping. Extensive ablation studies have been conducted on multiple datasets to evaluate the advantages of each component in our model. Additionally, physical experiments have been performed with natural language-driven human-robot interactions on a physical robot to validate the practicality of our approach.