Information Bottleneck (IB) based multi-view learning provides an information theoretic principle for seeking shared information contained in heterogeneous data descriptions. However, its great success is generally at...
详细信息
In Zermelo’s navigation problem (ZNP), classic solutions struggle with integrating obstacle avoidance constraints and adjusting the control law online in response to uncertain disturbances. Learning-based methods tha...
详细信息
In Zermelo’s navigation problem (ZNP), classic solutions struggle with integrating obstacle avoidance constraints and adjusting the control law online in response to uncertain disturbances. Learning-based methods that leverage prior knowledge offer a promising approach for achieving real-time optimal control. This paper proposes an end-to-end real-time optimal control strategy based on a deep neural network (DNN) with homotopy embedding for an underactuated unmanned surface vessel (USV) navigating dynamic environments with obstacles. The proposed framework leverages a bi-level structure, combining the advantages of DNN-based learning strategies and optimal control theory. To achieve this, optimal trajectories are generated by offline solving a nonlinear trajectory planning problem for the USV, considering environmental uncertainties and obstacle constraints. Under the Lipschitz continuity constraint, the DNN structure with homotopy parameter embedding utilizes a dataset incorporating prior knowledge of the navigation task to ensure stable learning. Consequently, the well-trained network can be effectively employed as an online controller to generate optimal feedback control based on the current state and varying disturbance intensities. Finally, numerical results are presented to demonstrate the viability and efficiency of the proposed control strategy in planning and steering the USV to its intended targets, showcasing robustness to disturbances from initial conditions and the environment, while also satisfying obstacle constraints.
Non-Gaussian statistical signal processing is be-coming increasingly significant in the current complex world. However, the variance of a non-Gaussian distribution may not exist, therefore many conventional methods wi...
详细信息
ISBN:
(数字)9781728159539
ISBN:
(纸本)9781728159546
Non-Gaussian statistical signal processing is be-coming increasingly significant in the current complex world. However, the variance of a non-Gaussian distribution may not exist, therefore many conventional methods will be considerably weakened, meaningless or even misleading. For example, the least-squares criterion may not be robust in the non-Gaussian environment, owing to the impulsive behaviors and outliers. In the paper, the statistical measures with fractional lower order moments (FLOM) are used instead of second-order moments to assess the control performance for non-Gaussian processes. The results show that FLOM is robust against outliers for the non-Gaussian signals. In the end, Hurst exponent fitting with FLOM and multifractal detrended fluctuation analysis (MFDFA) with FLOM are applied to the real data.
Correction for 'Bubble-enhanced ultrasonic microfluidic chip for rapid DNA fragmentation' by Lin Sun , , 2022, , 560-572, https://***/10.1039/D1LC00933H.
Correction for 'Bubble-enhanced ultrasonic microfluidic chip for rapid DNA fragmentation' by Lin Sun , , 2022, , 560-572, https://***/10.1039/D1LC00933H.
With the rapid development of intelligent vehicles, human drivers are sharing the control authority with the automation functionalities. The mutual understanding between the intelligent vehicle and the human driver is...
详细信息
With the rapid development of intelligent vehicles, human drivers are sharing the control authority with the automation functionalities. The mutual understanding between the intelligent vehicle and the human driver is the key to the efficient multi-agent teaming and the collaborative driving system. In this study, a two-stage learning framework for spatial-temporal driver lane change intention inference is developed. The two-stage learning framework contains two separate parts. First, an operational level driving behavior recognition system based on the deep convolutional neural network (CNN) is developed to recognize the driver behaviors, including mirror checking and normal driving. Then, based on the driver behavior features from the CNN, the sequential feature vector will be used for the construction of driver intention inference model based recurrent neural networks (RNN) and long short-term memory (LSTM). The proposed model achieved a state-of-the-art result on the driver lane change intention inference task. Based on the naturalistic driving dataset, the model achieved over 91% accuracy for lane change and lane-keeping intention prediction. The two-stage learning framework can significantly increase the model flexibility and accuracy, which makes it easier to be implemented and updated in intelligent vehicles.
Machine learning algorithms are commonly used to automate stock market trading. Crypto-currencies are novel digital assets that attracted investors all over the world. In this paper, we developed a neural network esti...
详细信息
Active prosthetic devices have been controlled using several methods such as Echo control, EMG signal based Position control and Finite State Machine (FSM) based Impedance/Compliance control. This manuscript proposes ...
Active prosthetic devices have been controlled using several methods such as Echo control, EMG signal based Position control and Finite State Machine (FSM) based Impedance/Compliance control. This manuscript proposes Virtual Constraint control of active prosthesis which obviates any need for the classification of EMG signals, identification of the gait phase for state switching and potentially avoids an exhaustive procedure for the tuning of impedance parameters. In this paper, a Discrete Fourier Transform (DFT) based Virtual Constraint control is presented to characterize the ankle-foot joint trajectory as a function of the human-inspired phase variable in a unified manner. An optimization-based algorithm is employed for the robust generation of continuously monotonic and linear phase variable for DFT based Virtual Constraint control. The results are generalized across various walking speeds for a specific user during the level ground walking.
This paper describes the use of mathematical modelling within flap system power driving unit for commercial aircraft. The modelling process is described from the architecture of flap system, the principle and mathemat...
详细信息
This paper describes the use of mathematical modelling within flap system power driving unit for commercial aircraft. The modelling process is described from the architecture of flap system, the principle and mathematical model of electrical motors and its control loop structure. The model of power driving unit and its external load system are developed in Simulink. Basic retraction and deployment function of flap has been performed, and the response of power driving unit under a typical jam failure of the actuator is provided. This work highlights the advantages of system modelling and simulation to support the load analysis of driveline components (e.g. actuators, torque shafts, etc.) within the conceptual design phase.
This paper describes the use of mathematical modelling within flap system power driving unit for commercial aircraft. The modelling process is described from the architecture of flap system, the principle and mathemat...
详细信息
In spired by the fact that a logarithm difference is the subtraction of two neighbor pixels, which may be a positive or negative numerical value, we divide a face local region into a positive region and a negative reg...
详细信息
In spired by the fact that a logarithm difference is the subtraction of two neighbor pixels, which may be a positive or negative numerical value, we divide a face local region into a positive region and a negative region, and the general logarithm difference model(GLDM) is developed by integrating positive and negative logarithm differences. Then, the multiscale logarithm difference edge-maps(MSLDE) [1] is employed as the test-bed, and the proposed GLDM is introduced into MSLDE to form General MSLDE(GMSLDE). Finally, the performance of GMSLDE is verified on the Extended Yale B and CMU PIE face databases with severe illumination variations. The experimental results indicate that the proposed GLDM can efficiently improve the performance of logarithm difference edge-map against severe illumination variations.
暂无评论