Direct current (DC) analysis lies at the heart of integrated circuit design in seeking DC operating points. Although pseudo-transient analysis (PTA) methods have been widely used in DC analysis in both industry and academia, their initial parameters and stepping strategy require expert knowledge and labor tuning to deliver efficient performance, which hinders their further applications. In this paper, we leverage the latest advancements in machine learning to deploy PTA with more efficient setups for different problems. More specifically, active learning, which automatically draws knowledge from other circuits, is used to provide suitable initial parameters for PTA solver, and then calibrate on-the-fly to further accelerate the simulation process using TD3-based reinforcement learning (RL). To expedite model convergence, we introduce dual agents and a public sampling buffer in our RL method to enhance sample utilization. To further improve the learning efficiency of the RL agent, we incorporate imitation learning to improve reward function and introduce supervised learning to provide a better dual-agent rotation strategy. We make the proposed algorithm a general out-of-the-box SPICE-like solver and assess it on a variety of circuits, demonstrating up to 3.10× reduction in NR iterations for the initial stage and 285.71× for the RL stage.

Sparse LU factorization is considered to be one of the most time-consuming components in circuit simulation, particularly when dealing with circuits of considerable size in the advanced process era. Sparse LU factorization can be expedited by utilizing the supernode structure, which partitions the matrix into dense sub-matrices, thereby improving computational performance by utilizing level-3 Basic Linear Algebra Subprograms (BLAS) General Matrix Multiplication (GEMM) operations. The sparse and irregular structure of circuit matrices often impedes the formation of supernodes or results in the formation of supernodes with many zero elements, which in turn poses challenges for exploiting GEMM operations. In this paper, by fully utilizing the density in sub-matrices and combining GEMM with the Dense-Sparse Matrix Multiplication (SpMM), we propose a density-aware adaptive matrix multiplication equipped with machine learning techniques to optimize performance of the most-time consuming matrix multiplication operator so as to accelerate the sparse LU factorization. Numerical experiment results show that among the 6 circuit matrices tested, the average performance of matrix multiplication in our algorithm can be improved by 5.35x (up to 9.35x) compared to the performance of using GEMM directly in Schur-complement updates. Compared with state-of-the-art solver SuperLU_DIST, our method shows a substantial performance improvement.

Relentless semiconductor scaling and ever increasing device integration have resulted in the exponentially growing size of the back-end design, which makes back-end simulation very time- and resource-consuming. With the success in the computer vision community, deep learning seems a promising alternative to assist the back-end simulation. However, unlike computer vision tasks, most back-end simulation problems are mathematically and physically well-defined, e.g., power delivery network sign off and post-layout circuit simulation. It then brings broad interests in the community where and how to deploy deep learning in the back-end simulation flows. This paper discusses a few challenges that the deployment of deep learning models in back-end simulation have to confront and the corresponding opportunities for future research.

DC analysis is the foundation for nonlinear electronic circuit simulation. Pseudo transient analysis (PTA) methods have gained great success among various continuation algorithms. However, PTA tends to be computationally intensive without careful tuning of parameters and proper stepping strategies. In this paper, we harness the latest advancing in machine learning to resolve these challenges simultaneously. Particularly, an active learning is leveraged to provide a fine initial solver environment, in which a TD3-based Reinforcement Learning (RL) is implemented to accelerate the simulation on the fly. The RL agent is strengthen with dual agents, priority sampling, and cooperative learning to enhance its robustness and convergence. The proposed algorithms are implemented in an out-of-the-box SPICElike simulator, which demonstrated a significant speedup: up to 3.1X for the initial stage and 234X for the RL stage.

Rapid and accurate estimation of the winter wheat leaf area index (LAI) is important for evaluating its growth and estimating yield. In this paper, Optimal Index (OI) was used to screen the best combination of hyperspectral bands in the flag stage and flowering period of wheat, and the LAI estimation model was constructed by Partial Least Square (PLS). The main results are as follows: The LAI estimation model based on the 614-774-794nm band combination is the best model for winter wheat flag stage (R2= 0.485, RMSE = 1.192, RV2= 0.682, RMSEV= 1.210); The LAI estimation model constructed by the 454-754-834nm band combination is the best model for winter wheat flowering (R2= 0.702, RMSE = 0.665, RV2= 0.810, RMSEV=0.468). The results show that it is feasible to use the optimal band combination as an independent variable to estimate the leaf area index of winter wheat, which can be used as a new method to monitor the growth of wheat.

The aboveground biomass of crops is an important index reflecting the growth status of crops. It is an important reference for achieving precise water and fertilizer management, yield monitoring and prediction. In this paper, the true height of the vegetation relative to the ground is obtained from the data preprocessing, the point cloud filtering and the normalization of the UAV radar data. The accuracy of biomass and height estimated by the average canopy height (Hcanopy) derived from canopy height model (CHM) is analyzed in a variety of resolution(5cm, 10cm, 15cm,20cm,25, 30cm,50cm, 100cm)respectively. The results suggest that the Hcanopyderived from CHM in 5cm resolution is highest correlated with biomass(R2=0.93, RMSE=698.90kg/ha), the Hcanopy derived from CHM in 50cm resolution is highest correlated with height (R2=0.96, RMSE=4cm). The crop growth parameters could be estimated based on UAV LiDAR across broad spatial scales.

Crop biomass is the basis of crop yield formation, and accurate biomass information is of great significance to ensure national food security. Taking winter wheat as the research object, the airborne radar data and wheat biomass information were used to study the difference of biomass inversion in the vertical distribution of laser point cloud at different scanning angles. The vertical distribution of point clouds is of great difference at 30 m above ground and a scan angle of ±60o. There is no correlation between the vertical distribution of point clouds and the scan angle of ±10oas a result of partial correlation analysis. Two lidar metrics (plot-level Hmea\mathfrakn and Dbelow) is strongly related to field-measured biomass (R2= 0.94,RMSE = 572.32kg/ha) at a scan angle of±10o.

Nitrogen is one of the important indices for evaluation of crop growth and output quality. Spectral reflectance of leaves and concurrent leaf nitrogen content parameters of samples were acquired in maize test. The top 10 different vegetation indices were chosen after ranking with VIP as the independent variable for estimating nitrogen content of leaf in maize. The leaf nitrogen content (LNC) estimation model with different vegetation indices can be built using the integrated model of variable importance projection (VIP) — partial least squares (PLS). The optimal model was selected by using Akaike's Information Criterion (AIC). The optimal model was validated by leave one out cross-validation (LOOCV) method. The decision coefficient (R2), root-mean-square error (RMSE) and relative error (RE) of the optimal model respectively were 0.73, 0.16 and 0; R2, RMSE and RE of maize by validating were 0.73, 0.19 and 0, respectively.

Structure-From-Motion (SFM) method is based on the same scene and different angles of the captured sequence of image, then calculate the feature points in the photogrammetric coordinate system of three-dimensional coordinates and camera parameters. SFM can directly generated orthophoto map just by captured overlapping images, but mosaicking accuracy has not to be verified. The purpose of this study is that verify the feasibility and accuracy of SFM method in UAV image mosaic. The process of UAV imagery mosaicking based on SFM method was elaborated, and the test image was mosaicked with UAV imagery processing software which based on SFM. The result: (1) UAV imagery mosaicking based on SFM algorithm has low accuracy on geographic positioning because of the low precision POS. But the distance\area measurement with high accuracy, the perimeter accuracy is above 96.6% and the area accuracy is above 93.2%. (2) The image had high accuracy after geometric correction using ground points. When 5 ground points were used, the mean value of absolute error was 0.60 m. The study showed: (1) the accuracy of perimeter and area can basically meet the accuracy requirements of distance\area measurement in agricultural applications. (2) the orthophoto map was rectified by ground control point can significantly improve the geo-location precision of the image.