The Standard WSO algorithm has several shortcomings, including uneven distribution of initial population, slow convergence speed, and weak global search ability. To address these issues, the present study proposes an improved War Strategy Optimization (WSO) based on hybrid strategy. To begin with, the initialization of the population was done using hypercube sampling. Additionally, diversification of the population during iteration process was achieved by adopting sine/cosine strategy, Cauthy mutation and backward learning strategy. Furthermore, to enhance capabilities in global search and local development, operator retention strategy from simulated annealing algorithm was employed. Finally, three test function optimization experiments were conducted which demonstrated that the proposed war strategy optimization algorithm based on hybrid strategy significantly improves both optimization accuracy and convergence speed.
In this study, a robust object recognition and command understanding system for a house tidying-up robot is proposed. The robot can understand the user's intentions by using a speech recognition system. When the user in-structs the tidying-up robot to tidy up an object using voice commands, the robot detects and recognizes the object. To detect and recognize multiple objects, we employ an instance segmentation method using deep learning. This method extracts the contour and shape of objects and generates the appropriate grasping posture of the robot arm. An experiment using the tidying-up robot is conducted to verify the effectiveness of the tidying-up system.
As an edge-cutting technology, Intelligent reflecting surface (IRS) can be well used in various fields. In this paper, we investigate to deploy double IRS in the dual-function radar and communication (DFRC) system, which we deploy double IRS to assist a base station (BS) with both radar sensing and communication capabilities with multiple single-antenna users. And we optimize the active beamforming at the base station and the passive beamforming of the IRS to maximize the achievable sum-rate of communication users. Because the optimization problems are non-convex, the fractional programming (FP), majorization-minimization (MM), and manifold optimization methods are used in an alternating algorithm to figure out these problems. Simulation results demonstrate the efficiency of the proposed algorithm.
This paper proposes a risk map generation method that considers the occupied space of objects and their characteristics. The objective is to guide visually impaired individuals safely to their intended destinations using a guide dog robot that can assist them in walking. The number of guide dogs in active service in Japan has been continuously declining, which has prompted the development of guide dog robots. The robot utilized in this study employs Mecanum wheels, the RoboSense RS-LiDAR-16 sensor, and Intel's RealSense Depth Camera D435 to scan the surrounding environment and measure distance up to 150 m. To prevent visually impaired individuals from entering spaces potentially occupied by objects, the three-dimensional spatial information of the objects is projected onto a two-dimensional map, and object recognition is performed to project the potential risks of objects onto the map. The generated risk map is used to path planning that considers the risk levels established according to object properties. The effectiveness is proven by experiments of guiding visually impaired individuals to destinations while avoiding potential occupied spaces.
The safe operation of the subway has great significance for the daily order of large cities. Under special circumstances, passengers and various objects may get caught between the subway train doors and the platform screen doors, which poses a safety hazard. We propose a method based on an RGBD dataset and a convolutional neural network binary classification model. By extracting features and using local features for normal and abnormal classification, we can detect in real-time whether there are people or objects caught between the subway double doors. Four sets of experiments were conducted using two classification models, with both RGBD and RGB datasets loaded, to demonstrate the advantages of using an RGBD dataset in improving accuracy. We also found a lightweight and high-accuracy model suitable for this application scenario to be run on edge devices, solving practical problems. By detecting foreign objects in the double-door gap, subway door anti-pinch measures can prevent passengers from getting caught between the subway train doors and the platform screen doors, ensuring passenger safety. This enhances the safety and reliability of the urban subway system, providing assurance for urban development.
