In order for a mobile robot to be able to effectively operate in complex, dynamic environments it must be capable of understanding both where and what the objects around them are. In this paper we introduce the semant...
详细信息
ISBN:
(纸本)9783030954598;9783030954581
In order for a mobile robot to be able to effectively operate in complex, dynamic environments it must be capable of understanding both where and what the objects around them are. In this paper we introduce the semantic probability hypothesis density (SPHD) filter, which allows robots to simultaneously track multiple classes of targets despite measurement uncertainty, including false positive detections, false negative detections, measurement noise, and target misclassification. The SPHD filter is capable of incorporating a different motion model for each type of target and of functioning in situations where the number of targets is unknown and time-varying. We demonstrate the efficacy of the SPHD filter via simulations with multiple target types containing both static and dynamic targets. We show that the SPHD filter performs better than a collection of PHD filters running in parallel, one for each target class.
In recent years, autonomous robotic peg-in-hole assembly has emerged as a prominent area of research. Due to the challenge of modeling and determining reward functions for pegin-hole assembly scenarios, imitation lear...
详细信息
ISBN:
(纸本)9798350388084;9798350388077
In recent years, autonomous robotic peg-in-hole assembly has emerged as a prominent area of research. Due to the challenge of modeling and determining reward functions for pegin-hole assembly scenarios, imitation learning is a more appropriate option as it enables learning policies from expert demonstrations without relying on environmental reward functions. However, traditional generative adversarial imitation learning (GaiL) methods are not efficient as they generate low-quality samples, resulting in high costs of interaction between the agent and the environment. This paper proposes a novel approach called hindsight transformation generative adversarial imitation learning (HT-GaiL), which employs hindsight experience to transform the samples generated by the generator into expert-like data. This expert-like data are then combined with expert demonstrations for adversarial network training. The proposed approach effectively addresses the issue of low utilization efficiency of generated samples, resulting in faster convergence of the algorithm. To validate the proposed algorithm, we constructed an autonomous peg-in-hole assembly platform using a six-degree-of-freedom manipulator. Comparative experiments were conducted to demonstrate the superior performance of the algorithm. The trained policies achieved success rates of 96%, 88%, and 61% for fit clearances of 1.12mm, 0.80mm, and 0.52mm, respectively, while maintaining assembly forces within 1.18N, 2.21N, and 9.10N, respectively.
The fresh food industry significantly depends on manual labor, which can make up to 40% of total production costs. Until now, implementing safe robotic automation for gently harvesting fresh produce has been difficult...
详细信息
The fresh food industry significantly depends on manual labor, which can make up to 40% of total production costs. Until now, implementing safe robotic automation for gently harvesting fresh produce has been difficult due to the complex and delicate nature of these tasks. The EU-funded SoftGrip Project aims to revolutionize the fresh food sector with technological advancements. By integrating artificial intelligence (ai) and robotic automation, it is possible to achieve gentle harvesting, enhance productivity, and lower labor costs for small and medium-sized European mushroom farms. The innovative smart soft gripper, designed to learn skills from expert harvesters through imitation learning, seeks to provide an economically feasible, scalable, and environmentally friendly solution, transforming the mushroom cultivation industry and the wider fruit market.
The aim of this paper is to develop the concept and a prototype of an intelligent mobile robotic platform that is integrated with nondestructive evaluation (NDE) capabilities for autonomous live inspection and repair....
详细信息
The aim of this paper is to develop the concept and a prototype of an intelligent mobile robotic platform that is integrated with nondestructive evaluation (NDE) capabilities for autonomous live inspection and repair. In many industrial environments, such as the application of power plant boiler inspection, human inspectors often have to perform hazardous and challenging tasks. There is a significant chance of injury, considering the confined spaces and limited visibility of the inspection environment and hazards such as pressurization and improper water levels. In order to provide a solution to eliminate these dangers, the concept of a new robotic system was developed and prototyped that is capable of autonomously sweeping the region to be inspected. The robot design contains systematic integration of components from robotics, NDE, and artificial intelligence (ai). A magnetic track system is used to navigate over the vertical steel structures required for examination. While moving across the inspection area, the robot uses an NDE sensor to acquire data for inspection and repair. This paper presents a design of a portable NDE scanning system based on eddy current array probes, which can be customized and installed on various mobile robot platforms. Machine learning methods are applied for semantic segmentation that will simultaneously localize and recognize defects without the need of human intervention. Experiments were conducted that show the NDE and repair capabilities of the system. Improvements in human safety and structural damage prevention, as well as lowering the overall costs of maintenance, are possible through the implementation of this robotic NDE system.
Obstacle contact detection is not commonly employed in autonomous robots, which mainly depend on avoidance algorithms, limiting their effectiveness in cluttered environments. Current contact-detection techniques suffe...
详细信息
Obstacle contact detection is not commonly employed in autonomous robots, which mainly depend on avoidance algorithms, limiting their effectiveness in cluttered environments. Current contact-detection techniques suffer from blind spots or discretized detection points, and rigid platforms further limit performance by merely detecting the presence of a collision without providing detailed feedback. To address these challenges, we propose an innovative contact sensor design that improves autonomous navigation through physical contact detection. The system features an elastic collision platform integrated with flex sensors to measure displacements during collisions. A neural network-based contact-detection algorithm converts the flex sensor data into actionable contact information. The collision system was validated with collisions through manual flights and autonomous contact-based missions, using sensor feedback for real-time collision recovery. The experimental results demonstrated the system's capability to accurately detect contact events and estimate collision parameters, even under dynamic conditions. The proposed solution offers a robust approach to improving autonomous navigation in complex environments and provides a solid foundation for future research on contact-based navigation systems.
暂无评论