Search

Question answering requiring discrete reasoning, e.g., arithmetic computing, comparison, and counting, over knowledge is a challenging task. In this paper, we propose UniRPG, a semantic-parsing-based approach advanced in interpretability and scalability, to perform unified discrete reasoning over heterogeneous knowledge resources, i.e., table and text, as program generation. Concretely, UniRPG consists of a neural programmer and a symbolic program executor, where a program is the composition of a set of pre-defined general atomic and higher-order operations and arguments extracted from table and text. First, the programmer parses a question into a program by generating operations and copying arguments, and then the executor derives answers from table and text based on the program. To alleviate the costly program annotation issue, we design a distant supervision approach for programmer learning, where pseudo programs are automatically constructed without annotated derivations. Extensive experiments on the TAT-QA dataset show that UniRPG achieves tremendous improvements and enhances interpretability and scalability compared with state-of-the-art methods, even without derivation annotation. Moreover, it achieves promising performance on the textual dataset DROP without derivations., Comment: Accepted to EMNLP 2022

Reasoning machine reading comprehension (R-MRC) aims to answer complex questions that require discrete reasoning based on text. To support discrete reasoning, evidence, typically the concise textual fragments that describe question-related facts, including topic entities and attribute values, are crucial clues from question to answer. However, previous end-to-end methods that achieve state-of-the-art performance rarely solve the problem by paying enough emphasis on the modeling of evidence, missing the opportunity to further improve the model's reasoning ability for R-MRC. To alleviate the above issue, in this paper, we propose an evidence-emphasized discrete reasoning approach (EviDR), in which sentence and clause level evidence is first detected based on distant supervision, and then used to drive a reasoning module implemented with a relational heterogeneous graph convolutional network to derive answers. Extensive experiments are conducted on DROP (discrete reasoning over paragraphs) dataset, and the results demonstrate the effectiveness of our proposed approach. In addition, qualitative analysis verifies the capability of the proposed evidence-emphasized discrete reasoning for R-MRC., Comment: 12 pages, 1 figure and 5 tables

The measuring of RCS has great meaning in the modern war environment and anti-stealth technology. This article introduces the process of RCS measuring. In order to get accurate RCS numerical, the calibration of the radar system is the first thing need to do. A standard metal ball with known RCS numerical hung by a balloon has been used as the calibration body in the external experiment, through contrast with the measuring data of the target object, the final radar RCS can be computed. The final experiment data show that the method given can achieve a satisfactory RCS numerical.

The embedded controller is small in size and powerful in computing power, and can quickly complete the related processing and calculation of flight attitude and track planning data. Wireless data transmission can realize a long-distance data transmission between the aircraft command center and the airborne control platform. Based on the data fusion algorithm of the Kalmanz filter, the data collected by multiple sensors can be integrated, which can effectively reduce the measurement noise amplitude. At the same time, the cumulative error of a single sensor is reduced. In order to solve the problems of ant colony algorithm in case it is easy to fall into the local extremum and the convergence speed is slow, an improved ant colony algorithm for 3D navigation of unmanned aerial vehicles is proposed for trace planning. This study divides the three-dimensional track planning into two parts based on the improved ant colony algorithm for two-dimensional plane planning and height planning. Geometric optimization methods to enhance the guidance of ant search are used. According to the distance and height constraints between the track point and the threat source, the altitude of the track points to plan the 3D track of the drone is calculated and adjusted. At the same time, the adaptive parameter adjustment method is used to improve the ant colony search ability and the interaction ability between individuals, and effectively get rid of the situation to avoid to falling into a local optimum. In addition, the index function is established and the path is smoothed. Simulation results show that the proposed improved algorithm cannot only safely avoid threats in the three-dimensional environment, but also has the ability to find the optimal solution and the convergence speed is better than the original algorithm.

Flow visualization of supersonic flow over a double ramp is achieved by the nano-tracer planar laser scattering (NPLS) technique in a supersonic low noise wind tunnel at Mach number 3.0. The fine structures of the boundary layer on the model, as well as the temporal and spatial evolutionary characteristics of the coherent structures, are revealed and investigated. High-spatiotemporal resolution images of the incoming boundary layer with laminar flow over the model are presented. Boundary layer transition occurs at the first ramp and the boundary layer at the second ramp has already been turbulent flow. The hairpin vortex and hairpin packet can be observed at the boundary layer of the first ramp. The angles of the hairpin vortex and hairpin packet are larger than that in a flat-plate because of adverse pressure gradient of the ramp. NPLS images of the incoming boundary layer with turbulent flow over the model are shown. The coherent structures at the first ramp are similar to the incoming boundary layer, which exhibit an obvious hierarchy. In contrast, the coherent structures at the second ramp exhibit different characteristics.

In a Mach 3.8 wind tunnel, both instantaneous and time-averaged flow structures of different scales around a blunt double-cone with or without supersonic film cooling were visualized via nano-tracer planar laser scattering (NPLS), which has a high spatiotemporal resolution. Three experimental cases with different injection mass flux rates were carried out. Many typical flow structures were clearly shown, such as shock waves, expansion fans, shear layers, mixing layers, and turbulent boundary layers. The analysis of two NPLS images with an interval of 5 μs revealed the temporal evolution characteristics of flow structures. With matched pressures, the laminar length of the mixing layer was longer than that in the case with a larger mass flux rate, but the full covered region was shorter. Structures like K—H (Kelvin—Helmholtz) vortices were clearly seen in both flows. Without injection, the flow was similar to the supersonic flow over a backward-facing step, and the structures were relatively simpler, and there was a longer laminar region. Large scale structures such as hairpin vortices were visualized. In addition, the results were compared in part with the schlieren images captured by others under similar conditions.