Frontiers in Robotics and AI最新文献

In this paper, we present a global reactive motion planning framework designed for robotic manipulators navigating in complex dynamic environments. Utilizing local minima-free circular fields, our methodology generates reactive control commands while also leveraging global environmental information from arbitrary configuration space motion planners to identify promising trajectories around obstacles. Furthermore, we extend the virtual agents framework introduced in Becker et al. (2021) to incorporate this global information, simulating multiple robot trajectories with varying parameter sets to enhance avoidance strategies. Consequently, the proposed unified robotic motion planning framework seamlessly combines global trajectory planning with local reactive control and ensures comprehensive obstacle avoidance for the entire body of a robotic manipulator. The efficacy of the proposed approach is demonstrated through rigorous testing in over 4,000 simulation scenarios, where it consistently outperforms existing motion planners. Additionally, we validate our framework's performance in real-world experiments using a collaborative Franka Emika robot with vision feedback. Our experiments illustrate the robot's ability to promptly adapt its motion plan and effectively avoid unpredictable movements by humans within its workspace. Overall, our contributions offer a robust and versatile solution for global reactive motion planning in dynamic environments.

There have been significant advances in robotics, conversational AI, and spoken dialogue systems (SDSs) over the past few years, but we still do not find social robots in public spaces such as train stations, shopping malls, or hospital waiting rooms. In this paper, we argue that early-stage collaboration between robot designers and SDS researchers is crucial for creating social robots that can legitimately be used in real-world environments. We draw from our experiences running experiments with social robots, and the surrounding literature, to highlight recurring issues. Robots need better speakers, a greater number of high-quality microphones, quieter motors, and quieter fans to enable human-robot spoken interaction in the wild. If a robot was designed to meet these requirements, researchers could create SDSs that are more accessible, and able to handle multi-party conversations in populated environments. Robust robot joints are also needed to limit potential harm to older adults and other more vulnerable groups. We suggest practical steps towards future real-world deployments of conversational AI systems for human-robot interaction.

Construction machines, for example cranes, excavators, or bulldozers, are widely diffused systems operating outdoors in harsh and dangerous environments, such as building sites, forests, and mines. Typically, construction machines require the on-site presence of highly skilled users to manage the complexity of their control and the high power of hydraulic actuation. Construction machines could benefit from recent developments of robot avatar technology that has demonstrated the viability of remotizing human physical activities, leveraging on intuitive interfaces and controls. Similar approaches could also improve the overall usability of construction machines, making them safer and accessible for untrained users. With this in mind, we developed a novel system for the remote control of cranes through intuitive and immersive interfaces. To validate the solution, we evaluated the experience of approximately 80 untrained users that remotely operated a crane during the 33rd Edition of Bauma, the world's leading fair for construction machines.

This work considers the problem of learning cooperative policies in multi-agent settings with partially observable and non-stationary environments without a communication channel. We focus on improving information sharing between agents and propose a new multi-agent actor-critic method called Multi-Agent Cooperative Recurrent Proximal Policy Optimization (MACRPO). We propose two novel ways of integrating information across agents and time in MACRPO: First, we use a recurrent layer in the critic's network architecture and propose a new framework to use the proposed meta-trajectory to train the recurrent layer. This allows the network to learn the cooperation and dynamics of interactions between agents, and also handle partial observability. Second, we propose a new advantage function that incorporates other agents' rewards and value functions by controlling the level of cooperation between agents using a parameter. The use of this control parameter is suitable for environments in which the agents are unable to fully cooperate with each other. We evaluate our algorithm on three challenging multi-agent environments with continuous and discrete action spaces, Deepdrive-Zero, Multi-Walker, and Particle environment. We compare the results with several ablations and state-of-the-art multi-agent algorithms such as MAGIC, IC3Net, CommNet, GA-Comm, QMIX, MADDPG, and RMAPPO, and also single-agent methods with shared parameters between agents such as IMPALA and APEX. The results show superior performance against other algorithms. The code is available online at https://github.com/kargarisaac/macrpo.

In recent years, providing additional visual feedback about the interaction forces has been found to offer benefits to haptic-assisted teleoperation. However, there is limited insight into the effects of the design of force feedback-related visual cues and the type of visual display on the performance of teleoperation of robotic arms executing industrial tasks. In this study, we provide new insights into this interaction by extending these findings to the haptic assistance teleoperation of a simulated robotic arm in a virtual environment, in which the haptic assistance is comprised of a set of virtual fixtures. We design a novel method for providing visual cues about the interaction forces to complement the haptic assistance and augment visual feedback in virtual reality with a head-mounted display. We evaluate the visual cues method and head-mounted display method through human factors experiments in a teleoperated dross removal use case. The results show that both methods are beneficial for task performance, each of them having stronger points in different aspects of the operation. The visual cues method was found to significantly improve safety in terms of peak collision force, whereas the head-mounted display additionally improves the performance significantly. Furthermore, positive scores of the subjective analysis indicate an increased user acceptance of both methods. This work provides a new study on the importance of visual feedback related to (interaction) forces and spatial information for haptic assistance and provides two methods to take advantage of its potential benefits in the teleoperation of robotic arms.

Background: Accurate histological analysis is crucial for confirming intracerebral neoplasia due to the diverse array of potential diagnoses presented by imaging. In the realm of biopsy techniques, the use of robot-based systems is on the rise, primarily owing to their heightened targeting accuracy. The objective of this study was to elucidate the practicality, learning curve and workflow associated with robot-guided biopsies upon their introduction to a neurosurgical centre.

Materials and methods: In March 2022, the neurosurgical department at our institution adopted the Medtronic Stealth AutoGuide™ cranial robotic guidance platform, a miniaturized robotic guidance device designed for stereotactic procedures. Four experienced neurosurgeons underwent training with the Stealth AutoGuide™ system, after which 51 consecutive biopsies were performed. The evaluation encompassed entry and target accuracy, preoperative setup time, time required for the biopsy procedure, and overall operating time. Statistical analysis was conducted to identify any differences between the initial 26 and subsequent sets of 25 patients, with the aim of identifying changes in workflow.

Results: The study included all patients necessitating a diagnostic biopsy for intracerebral tumours, with only one patient excluded due to the inaccessibility of the intended target point caused by tumour calcification. Notably, there were no significant differences between the first 25 and last 26 patients in the median time from incision to the first biopsy (overall: 11.5 min, IQR 9.03-15.0), the procedure time (overall: 30.0 min, IQR 23.5-46.5), median accuracy at entry (overall: 2.05 mm, IQR 0.8-3.8), or target point (overall: 2.2 mm (IQR 1.6-3.7). However, a significant reduction in robot setup time was observed between the two groups, median 2.69 min versus 1.17 min, respectively (p ≤ 0.001).

Conclusion: The deployment of the robotic biopsy system, Stealth AutoGuide™, showed high feasibility, steep learning curve due to uncomplicated technical handling using our standardized operative workflow. Therefore, even in prone position a high diagnostic yield was achieved. Overall, the Stealth AutoGuide™ system facilitated biopsies in traditionally challenging regions with concise procedure time and surgeon-independent high accuracy.

Health is a multidimensional phenomenon encompassing physical, mental, social, and spiritual aspects. Therefore, it is only logical that good care should be holistic and include all these dimensions. However, when examining the current field of health and care robotics, the spiritual aspect is notably neglected. As a result, current health and care robots cannot provide holistic care. This paper argues that this neglect should be addressed, and, drawing on the emerging field of spiritual robotics, that spiritual aspects should receive greater attention when considering, developing, or deploying health and care robots. We also propose guidelines for equipping health and care robots with the necessary spiritual capabilities.

Research on social assistive robots in education faces many challenges that extend beyond technical issues. On one hand, hardware and software limitations, such as algorithm accuracy in real-world applications, render this approach difficult for daily use. On the other hand, there are human factors that need addressing as well, such as student motivations and expectations toward the robot, teachers' time management and lack of knowledge to deal with such technologies, and effective communication between experimenters and stakeholders. In this paper, we present a complete evaluation of the design process for a robotic architecture targeting teachers, students, and researchers. The contribution of this work is three-fold: (i) we first present a high-level assessment of the studies conducted with students and teachers that allowed us to build the final version of the architecture's module and its graphical interface; (ii) we present the R-CASTLE architecture from a technical perspective and its implications for developers and researchers; and, finally, (iii) we validated the R-CASTLE architecture with an in-depth qualitative analysis with five new teachers. Findings suggest that teachers can intuitively import their daily activities into our architecture at first glance, even without prior contact with any social robot.

To address the problems of the labeling curved surfaces vegetable with long label, such as the label wrinkled and the easy detachment, a cam-elliptical gear combined labeling mechanism with an improved hypocycloid trajectory is proposed. Provide the process of the mechanism, and establish a kinematic model of the mechanism. In order to improve the motion performances of the cam-elliptical gear combined labeling mechanism and avoid labels damage, the NSGA-II algorithm is used to optimize the parameters of the mechanism, resulting in 80 sets of Pareto solutions. The entropy weight TOPSIS method is applied as a quadratic optimization to select an optimal solution from the 80 sets of Pareto solutions and obtain the optimized parameters of the mechanism. A comparative study is conducted with an elliptical-circular planetary gear mechanism using the hypocycloid trajectory. The results show that the improved mechanism reduces the maximum velocity by 7%, the maximum and minimum accelerations by 2% and 18%. After the quadratic optimization the distance error of the center point of suction cup and the labeling point is reduced from 1.3 mm to 0.12 mm, and the velocity during labeling and taking position is reduced from 0.10770 m s^-1 to 0.0037 m s^-1. The correctness of the proposed method is validated through simulation studies and experiments. This research provides a theoretical basis for the design and optimization of long label and curved surface labeling mechanism for vegetables.