Search

Proximal policy optimization (PPO) is a deep reinforcement learning algorithm based on the actor–critic (AC) architecture. In the classic AC architecture, the Critic (value) network is used to estimate the value function while the Actor (policy) network optimizes the policy according to the estimated value function. The efficiency of the classic AC architecture is limited due that the policy does not directly participate in the value function update. The classic AC architecture will make the value function estimation inaccurate, which will affect the performance of the PPO algorithm. For improvement, we designed a novel AC architecture with policy feedback (AC-PF) by introducing the policy into the update process of the value function and further proposed the PPO with policy feedback (PPO-PF). For the AC-PF architecture, the policy-based expected (PBE) value function and discount reward formulas are designed by drawing inspiration from expected Sarsa. In order to enhance the sensitivity of the value function to the change of policy and to improve the accuracy of PBE value estimation at the early learning stage, we proposed a policy update method based on the clipped discount factor. Moreover, we specifically defined the loss functions of the policy network and value network to ensure that the policy update of PPO-PF satisfies the unbiased estimation of the trust region. Experiments on Atari games and control tasks show that compared to PPO, PPO-PF has faster convergence speed, higher reward, and smaller variance of reward. [ABSTRACT FROM AUTHOR]

This brief is concerned with the event-triggered control of linear systems with triggered dual channels. The system outputs and control inputs are respectively driven by two different triggering mechanisms. An integral controller with nonlinear compensation terms is established to attenuate the potential data jitter and reduce the unnecessary transmissions. The uniformly ultimate boundedness of the resultant closed-loop system is established by means of Simpson rule, Finsler lemma and the Lyapunov stability theory and the controller gain is solved in terms of linear matrix inequality. The validity of the proposed scheme is verified by a numerical example. [ABSTRACT FROM AUTHOR]

Input scale plays an important role in modern detection frameworks, and an optimal training scale for images exists empirically. However, the optimal one usually cannot be reached in facing extremely large images under the memory constraint. In this study, we explore the scale effect inside the object detection pipeline and find that feature upsampling with the introduction of high-resolution information benefits the detection. Compared with direct input upscaling, feature upsampling trades a small performance loss for a large amount of memory savings. From these observations, we propose a self-supervised feature augmentation network, which takes downsampled images as inputs and aims to generate comparable features with the ones when feeding upscaled images to networks. We present a guided feature upsampling module, which takes downsampled images as inputs, to learn upscaled feature representations with the supervision of real large features acquired from upscaled images. In a self-supervised learning manner, we can introduce detailed information of images to the network. For an efficient feature upsampling, we design a residualized sub-pixel convolution block based on a sub-pixel convolution layer, which involves considerable information in upsampling process. Experiments on Mapillary Vistas Dataset (MVD), Cityscapes, and COCO are conducted to demonstrate the effectiveness of our method. On the MVD and Cityscapes detection benchmarks, in which the images are extremely large, our method surpasses current approaches. On COCO, the proposed method obtains comparable results to existing methods but with higher efficiency. [ABSTRACT FROM AUTHOR]

This paper describes a new method for determining phase angle errors of shunts with rated currents at 50 and 100 A at frequencies from 25 to 100 kHz. The method is based on the use of a three-branch binary inductive current divider to measure the phase angle errors of high current shunts against a phase angle reference standard with only one step. An inductive high current shunt has been designed using a 100:1 current transformer connected in parallel with a 1-\Omega shunt. The phase angle errors of resistive high current shunts and the inductive high current shunt have been measured. The level dependence in phase angle errors of three different types of high current shunts has been evaluated at currents from 1 to 100 A. The measurement results and standard uncertainties of the phase angle errors of high current shunts are reported. [ABSTRACT FROM PUBLISHER]

We report on the measurement techniques and results of the equivalent inductance of current shunts at frequencies from 50 to 200 kHz. With a three-branch binary inductive current divider and difference detection transformer, the equivalent inductance of the cagelike shunts was measured against a set of four-terminal resistors whose time constants were accurately determined. The verification experiments for the systematic errors of the measurement setup and the phase comparator with the current ratio of 2 : 1 are also presented. [ABSTRACT FROM PUBLISHER]

This paper describes a new method to determine the phase angle errors of ac shunts by measuring the inductance and distributed capacitance. A 1-\Omega shunt of coaxial design has been developed as the time constant standard. A coaxial inductor with identical structure as the time constant standard has been developed. A four-terminal mutual inductor has also been built for the measurement of the inductance of the time constant standard. The method is based on the use of a binary inductive current divider to compare the inductor with the mutual inductor. The standard uncertainty of the inductance measurement of the time constant standard is within 28 pH at frequencies up to 200 kHz. The expanded uncertainty of the phase angle errors is from 0.4 \mu\rad at 1 kHz to 80.0 \mu\rad at 200 kHz at 1 A. [ABSTRACT FROM PUBLISHER]

This letter presents a constraint relaxation optimal power flow (OPF) model to tackle the issues when traditional OPF is infeasible under large variations such as wind generation output. In this model, the original hard constraints are relaxed into soft constraints and the objective function is adjusted for the cost of constraint violations. To guarantee the equivalence to the original OPF model when there are feasible solutions, an exact penalty function method is introduced to justify the selection of penalty factor of constraint violations. By solving an optimization problem, the lower bound of the proper penalty factor is obtained. The results of a 6-bus test system show that the proposed method achieves the same solution when the original OPF has feasible region, and an optimal solution can be obtained with minimum constraint violation when original OPF has no feasible region. Lastly, three large IEEE systems are tested to verify the effectiveness of proposed method. [ABSTRACT FROM PUBLISHER]