IET Cybersystems and Robotics最新文献

Speech interaction is an important means of robot interaction. With the rapid development of deep learning, end-to-end speech synthesis methods based on this technique have gradually become mainstream. Chinese deep learning-based speech synthesis techniques suffer from problems such as unstable synthesised speech, poor naturalness and poor personalised speech synthesis, which do not satisfy some practical application scenarios. Hence, an F-MelGAN model is adopted to improve the performance of Chinese speech synthesis. A post-processing network is used to refine the Mel-spectrum predicted by the decoder and alleviate the Mel-spectrum distortion phenomenon. A phoneme-level and sentence-level combined module is proposed to model the personalised style of speakers. A combination of an acoustic conditioning network, speaker encoder network GCNet and feedback-constrained training is proposed to solve the problem of poor personalised speech synthesis and achieve personalised speech customisation in Chinese. Experimental results show that the whole model can generate high-quality speech with high speaker similarity for both speakers that appear in the training process and speakers that never appear in the training process.

With the rapid development of artificial intelligence technology, commercial robots have gradually entered our daily lives. In order to promote product dissemination, shopping guide robots are a new service options of commerce platforms that use tag recommendation systems to identify users' intentions. A large number of applications combine user historical tagging information with the multi-round dialogue ability of shopping guide robots to help users efficiently search for and retrieve products of interest. Recently, tensor decomposition methods have become a common approach for modelling entity interaction relationships in tag recommendation systems. However, due to the sparsity of data, these methods only consider low-order information of entities, making it difficult to capture the higher-order collaborative signals among entities. Recommendation methods by autoencoders can effectively extract abstract feature representations while they only focus on the two-dimensional relationship between users and items, ignoring the interaction relationship among users, items and tags in real complex recommendation scenarios. The authors focus on modelling the similarity relationship among entities and propose a method called deep feature fusion tag (DFFT) based on the deep feature fusion of stacked denoising autoencoders. This method can extract high-order information with different embedding dimensions and fuse them in a unified framework. To extract robust feature representations, the authors inject random noise (mask-out/drop-out noise) into the tag information corresponding to users and items to generate corrupted input data, and then utilise autoencoders to encode the interaction relationship among entities. To further obtain the interaction relationship with different dimensions, different encoding layers are stacked and combined to produce a better expanded model which can reinforce each other. Finally, a decoding component is used to reconstruct the original input data. According to the experimental results on two common datasets, the proposed DFFT method outperforms other baselines in terms of the F1@N, NDCG@N and Recall@N evaluation metrics.

Localization is a core problem in mobile robot navigation. Simultaneous localization and mapping (SLAM) costs much for an unmanned aerial vehicle (UAV). This research aims to design an orthogonal laser scan device for localization and to save computation costs. Based on disturbance analysis, residual influences on sensor state are quantitative, and they are related to uncertainty and sensitivity. This research applied the residual selection method to a UAV. The feature point detection utilises multi-scale and Gaussian model fitting techniques to guarantee true positives. The map is represented by Gaussian Mixture Models (GMM) with lower memory costs. The orthogonal laser scan device is composed and placed on a UAV for real-time three-dimensional localization, whose errors are at the centimeter level.

When a person's neuromuscular system is affected by an injury or disease, Activities-for-Daily-Living (ADL), such as gripping, turning, and walking, are impaired. Electroencephalography (EEG) and Electromyography (EMG) are physiological signals generated by a body during neuromuscular activities embedding the intentions of the subject, and they are used in Brain–Computer Interface (BCI) or robotic rehabilitation systems. However, existing BCI or robotic rehabilitation systems use signal classification technique limitations such as (1) missing temporal correlation of the EEG and EMG signals in the entire window and (2) overlooking the interrelationship between different sensors in the system. Furthermore, typical existing systems are designed to operate based on the presence of dominant physiological signals associated with certain actions; (3) their effectiveness will be greatly reduced if subjects are disabled in generating the dominant signals. A novel classification model, named BIOFIS is proposed, which fuses signals from different sensors to generate inter-channel and intra-channel relationships. It explores the temporal correlation of the signals within a timeframe via a Long Short-Term Memory (LSTM) block. The proposed architecture is able to classify the various subsets of a full-range arm movement that performs actions such as forward, grip and raise, lower and release, and reverse. The system can achieve 98.6% accuracy for a 4-way action using EEG data and 97.18% accuracy using EMG data. Moreover, even without the dominant signal, the accuracy scores were 90.1% for the EEG data and 85.2% for the EMG data. The proposed mechanism shows promise in the design of EEG/EMG-based use in the medical device and rehabilitation industries.

To ensure that the robot can follow the planned trajectory, smooth switching between swinging legs and a smooth transition of motion process is realised. The previous motion planning work is analysed, and a method for improving the optimisation objective function and constraint conditions is proposed to eliminate the sudden change of acceleration and reduce the peak value of acceleration change. This method eliminates the impact phenomenon in the motor drive process and reduces the motor drive energy consumption, thus ensuring the smooth and consistent movement of the robot. The results show that the improved optimisation method has a better motion effect than the previous approach in terms of centre of mass motion speed, trajectory fitting and body posture change, and realises more robust motion of quadruped robots in a senseless state.

Modern agricultural tractors are complex systems, in which multiple physical (and technological) domains interact to reach a wide set of competing goals, including work operational performance and energy efficiency. This complexity translates to the dynamic, multi-domain simulation models implemented to serve as digital twins, for rapid prototyping and effective pre-tuning, prior to bench and on-field testing. Consequently, a suitable simulation framework should have the capability to focus both on the vehicle as a whole and on individual subsystems. For each of the latter, multiple options should be available, with different levels of detail, to properly address the relevant phenomena, depending on the specific focus, for an optimal balance between accuracy and computation time. The methodology proposed here by the authors is based on the lumped parameter approach and integrates the models for the following subsystems in a modular context: internal combustion engine, hydromechanical transmission, vehicle body, and tyre–soil interaction. The model is completed by a load cycle module that generates stimulus time histories to reproduce the work load under real operating conditions. Traction capability is affected by vertical load on the wheels, which is even more relevant if the vehicle is travelling on an uncompacted soil and subject to a variable drawbar pull force as it is when ploughing. The vertical load is, in turn, heavily affected by vehicle dynamics, which can be accurately modelled via a full multibody implementation. The presented lumped parameter model is intended as a powerful simulation tool to evaluate tractor performance, both in terms of fuel consumption and traction dynamics, by considering the cascade phenomena from the wheel–ground interaction to the engine, passing through the dynamics of vehicle bodies and their mass transfer. Its capabilities and numerical results are presented for the simulation of a realistic ploughing operation.

Considering the wheeled mobile robot (WMR) tracking problem with velocity saturation, we developed a data-driven iterative learning double loop control method with constraints. First, the authors designed an outer loop controller to provide virtual velocity for the inner loop according to the position and pose tracking error of the WMR kinematic model. Second, the authors employed dynamic linearisation to transform the dynamic model into an online data-driven model along the iterative domain. Based on the measured input and output data of the dynamic model, the authors identified the parameters of the inner loop controller. The authors considered the velocity saturation constraints; we adjusted the output velocity of the WMR online, providing effective solutions to the problem of velocity saltation and the saturation constraint in the tracking process. Notably, the inner loop controller only uses the output data and input of the dynamic model, which not only enables the reliable control of WMR trajectory tracking, but also avoids the influence of inaccurate model identification processes on the tracking performance. The authors analysed the algorithm's convergence in theory, and the results show that the tracking errors of position, angle and velocity can converge to zero in the iterative domain. Finally, the authors used a simulation to demonstrate the effectiveness of the algorithm.

Robotic fish actuated by smart materials has attracted extensive attention and has been widely used in many applications. In this study, a robotic fish actuated by dielectric elastomer (DE) films is proposed. The tensile behaviours of DE film VHB4905 are studied, and the Ogden constitutive equation is employed to describe the stress-strain behaviour of the DE film. The fabrication processes of the robotic fish, including pre-stretching treatment of the DE films, electrode coating with carbon paste, and waterproof treatment, are illustrated in detail. The dynamic response of the fabricated DE actuators under different excitation voltages is tested based on the experimental setup. Experimental results show that the first-order natural frequencies of the obtained DE actuator in air is 4.05 Hz. Finally, the swimming performances of the proposed robotic fish at different driving levels are demonstrated, and it achieves an average swimming speed of 20.38 mm/s, with a driving voltage of 5kV at 0.8 Hz.

In this study, a practical adaptive control scheme is proposed for the trajectory tracking of an unmanned surface vehicle via the characteristic modelling approach. Therefore, accurate tracking control can be achieved in the presence of unknown time-varying model parameters and environmental disturbances. The control scheme comprises a trajectory guidance module based on the virtual target approach and a tracking control module designed by characteristic modelling theory. Firstly, the ideal control commands of the yaw speed and surge speed are generated using the position errors between the vehicle and the virtual target. Then, a second-order characteristic model for the heading and surge speed channel is developed. The parameters of the model are updated by a real-time parameter identification algorithm. Based on this model, an integrated adaptive control law is designed which consists of golden-section control, feed-forward control and integral control. Finally, the development processes of the vehicle platform and the control algorithms are described, and the results of simulation and field experiments are presented and discussed.