Frontiers in Neurorobotics最新文献

Autonomous driving technology has garnered significant attention due to its potential to revolutionize transportation through advanced robotic systems. Despite optimistic projections for commercial deployment, the development of sophisticated autonomous driving systems remains largely experimental, with the effectiveness of neurorobotics-based decision-making and planning algorithms being crucial for success. This paper delivers a comprehensive review of decision-making and planning algorithms in autonomous driving, covering both knowledge-driven and data-driven approaches. For knowledge-driven methods, this paper explores independent decision-making systems, including rule based, state transition based, game-theory based methods and independent planing systems including search based, sampling based, and optimization based methods. For data-driven methods, it provides a detailed analysis of machine learning paradigms such as imitation learning, reinforcement learning, and inverse reinforcement learning. Furthermore, the paper discusses hybrid models that amalgamate the strengths of both data-driven and knowledge-driven approaches, offering insights into their implementation and challenges. By evaluating experimental platforms, this paper guides the selection of appropriate testing and validation strategies. Through comparative analysis, this paper elucidates the advantages and disadvantages of each method, facilitating the design of more robust autonomous driving systems. Finally, this paper addresses current challenges and offers a perspective on future developments in this rapidly evolving field.

Aiming at the problems of slow network convergence, poor reward convergence stability, and low path planning efficiency of traditional deep reinforcement learning algorithms, this paper proposes a BiLSTM-D3QN (Bidirectional Long and Short-Term Memory Dueling Double Deep Q-Network) path planning algorithm based on the DDQN (Double Deep Q-Network) decision model. Firstly, a Bidirectional Long Short-Term Memory network (BiLSTM) is introduced to make the network have memory, increase the stability of decision making and make the reward converge more stably; secondly, Dueling Network is introduced to further solve the problem of overestimating the Q-value of the neural network, which makes the network able to be updated quickly; Adaptive reprioritization based on the frequency penalty function is proposed. Experience Playback, which extracts important and fresh data from the experience pool to accelerate the convergence of the neural network; finally, an adaptive action selection mechanism is introduced to further optimize the action exploration. Simulation experiments show that the BiLSTM-D3QN path planning algorithm outperforms the traditional Deep Reinforcement Learning algorithm in terms of network convergence speed, planning efficiency, stability of reward convergence, and success rate in simple environments; in complex environments, the path length of BiLSTM-D3QN is 20 m shorter than that of the improved ERDDQN (Experience Replay Double Deep Q-Network) algorithm, the number of turning points is 7 fewer, the planning time is 0.54 s shorter, and the success rate is 10.4% higher. The superiority of the BiLSTM-D3QN algorithm in terms of network convergence speed and path planning performance is demonstrated.

Human skeleton-based action recognition is an important task in the field of computer vision. In recent years, masked autoencoder (MAE) has been used in various fields due to its powerful self-supervised learning ability and has achieved good results in masked data reconstruction tasks. However, in visual classification tasks such as action recognition, the limited ability of the encoder to learn features in the autoencoder structure results in poor classification performance. We propose to enhance the encoder's feature extraction ability in classification tasks by leveraging the latent space of variational autoencoder (VAE) and further replace it with the latent space of vector quantized variational autoencoder (VQVAE). The constructed models are called SkeletonMVAE and SkeletonMVQVAE, respectively. In SkeletonMVAE, we constrain the latent variables to represent features in the form of distributions. In SkeletonMVQVAE, we discretize the latent variables. These help the encoder learn deeper data structures and more discriminative and generalized feature representations. The experiment results on the NTU-60 and NTU-120 datasets demonstrate that our proposed method can effectively improve the classification accuracy of the encoder in classification tasks and its generalization ability in the case of few labeled data. SkeletonMVAE exhibits stronger classification ability, while SkeletonMVQVAE exhibits stronger generalization in situations with fewer labeled data.

Introduction: Object perception, particularly material detection, is predominantly performed through texture recognition, which presents significant limitations. These methods are insufficient to distinguish between different materials with similar surface roughness, and noise caused by tactile movements affects the system performance.

Methods: This paper presents a straightforward, impact-based approach to identifying materials, utilizing the cantilever beam mechanism in the UR5e robot's artificial finger. To detect object material, an elastic metal sheet was fixed to a load cell with an accelerometer and a metal appendage positioned above and below its free end, respectively. After recording the damping force signal and vibration data from the load cell and accelerometer caused by the metal appendage's impact, features such as vibration amplitude, damping time, wavelength, and force amplitude were retrieved. Three machine-learning techniques were then used to classify the objects' materials according to their damping rates. Data clustering was performed using the deflection of the cantilever beam to boost classification accuracy.

Results and discussion: Online object materials detection shows an accuracy of 95.46% in a study of ten objects [metals (steel, cast iron), plastics (foam, compressed plastic), wood, silicon, rubber, leather, brick and cartoon]. This method overcomes the limitations of the tactile approach and has the potential to be used in industrial robots.

Introduction: To enhance the detection of litchi fruits in natural scenes, address challenges such as dense occlusion and small target identification, this paper proposes a novel multimodal target detection method, denoted as YOLOv5-Litchi.

Methods: Initially, the Neck layer network of YOLOv5s is simplified by changing its FPN+PAN structure to an FPN structure and increasing the number of detection heads from 3 to 5. Additionally, the detection heads with resolutions of 80 × 80 pixels and 160 × 160 pixels are replaced by TSCD detection heads to enhance the model's ability to detect small targets. Subsequently, the positioning loss function is replaced with the EIoU loss function, and the confidence loss is substituted by VFLoss to further improve the accuracy of the detection bounding box and reduce the missed detection rate in occluded targets. A sliding slice method is then employed to predict image targets, thereby reducing the miss rate of small targets.

Results: Experimental results demonstrate that the proposed model improves accuracy, recall, and mean average precision (mAP) by 9.5, 0.9, and 12.3 percentage points, respectively, compared to the original YOLOv5s model. When benchmarked against other models such as YOLOx, YOLOv6, and YOLOv8, the proposed model's AP value increases by 4.0, 6.3, and 3.7 percentage points, respectively.

Discussion: The improved network exhibits distinct improvements, primarily focusing on enhancing the recall rate and AP value, thereby reducing the missed detection rate which exhibiting a reduced number of missed targets and a more accurate prediction frame, indicating its suitability for litchi fruit detection. Therefore, this method significantly enhances the detection accuracy of mature litchi fruits and effectively addresses the challenges of dense occlusion and small target detection, providing crucial technical support for subsequent litchi yield estimation.

Slip detection is to recognize whether an object remains stable during grasping, which can significantly enhance manipulation dexterity. In this study, we explore slip detection for five-finger robotic hands being capable of performing various grasp types, and detect slippage across all five fingers as a whole rather than concentrating on individual fingertips. First, we constructed a dataset collected during the grasping of common objects from daily life across six grasp types, comprising more than 200 k data points. Second, according to the principle of deep double descent, we designed a lightweight universal slip detection convolutional network for different grasp types (USDConvNet-DG) to classify grasp states (no-touch, slipping, and stable grasp). By combining frequency with time domain features, the network achieves a computation time of only 1.26 ms and an average accuracy of over 97% on both the validation and test datasets, demonstrating strong generalization capabilities. Furthermore, we validated the proposed USDConvNet-DG in real-time grasp force adjustment in real-world scenarios, showing that it can effectively improve the stability and reliability of robotic manipulation.

For the researches of cooperative control scheme for multirobot systems, this paper sets up an experimental platform based on dobot robots, which can be used to perform physical experiments to verify related schemes. A distributed scheme is proposed to achieve cooperative control for multirobot systems. Simulation results prove the effectiveness of the distributed scheme. Then, the experimental platform based on dobot robots is built to verify the proposed scheme. Specifically, a computer sends data to the microcontroller inside the host through WiFi communication, then the host distributes data to the slaves. Finally, the physical experiment of related schemes is performed on the experimental platform. Comparing the simulations with the physical experiments, the task is successfully completed on this experimental platform, which proves the effectiveness of the scheme and the feasibility of the platform. The experimental platform developed in this paper possesses the capability to validate various schemes and exhibits strong expandability and practicality.

Time angle of arrival (AoA) and time difference of arrival (TDOA) are two widely used methods for solving dynamic signal source localization (DSSL) problems, where the position of a moving target is determined by measuring the angle and time difference of the signal's arrival, respectively. In robotic manipulator applications, accurate and real-time joint information is crucial for tasks such as trajectory tracking and visual servoing. However, signal propagation and acquisition are susceptible to noise interference, which poses challenges for real-time systems. To address this issue, a noise-immune zeroing neural dynamics (NIZND) model is proposed. The NIZND model is a brain-inspired algorithm that incorporates an integral term and an activation function into the traditional zeroing neural dynamics (ZND) model, designed to effectively mitigate noise interference during localization tasks. Theoretical analysis confirms that the proposed NIZND model exhibits global convergence and high precision under noisy conditions. Simulation experiments demonstrate the robustness and effectiveness of the NIZND model in comparison to traditional DSSL-solving schemes and in a trajectory tracking scheme for robotic manipulators. The NIZND model offers a promising solution to the challenge of accurate localization in noisy environments, ensuring both high precision and effective noise suppression. The experimental results highlight its superiority in real-time applications where noise interference is prevalent.

Dexterous hands play vital roles in tasks performed by humanoid robots. For the first time, we quantify the correlation between design variables and the performance of 65 dexterous hands using Cramér's V. Comprehensive cross-correlation analysis quantitatively reveals how the performance, such as speed, weight, fingertip force, and compactness are related to the design variables including degrees of freedom (DOF), structural form, driving form, and transmission mode. This study shows how various design parameters are coupled inherently, leading to compromise in performance metrics. These findings provide a theoretical basis for the design of dexterous hands in various application scenarios and offer new insights for performance optimization.

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.