Learning to navigate in an unknown environment is a crucial capability of mobile robot. Conventional method for robot navigation consists three steps, involving localization, map building and path planning. However, most the conventional methods rely on obstacle map, dont have ability autonomous learning. In contrast traditional approach, we propose end-to-end approach this paper using deep reinforcement learning robots environment. Based dueling network architectures (Dueling DQN) with...

Vision-based simultaneous localization and mapping (VSLAM) which uses visual sensor to make a robot locate itself in an unknown environment while simultaneously construct map of the environment. With continuous development computer vision robotics, VSLAM has become supporting technology for popular fields such as unmanned aerial vehicle, virtual reality driving. In this paper, classical framework SLAM is introduced briefly. On basis, key technologies latest research progress from indirect...

Low-cost microelectro mechanical systems (MEMS)-based inertial measurement unit (IMU) measurements are usually affected by inaccurate scale factors, axis misalignments, and g-sensitivity errors. These errors may significantly influence the performance of visual-inertial methods. In this paper, we propose an online IMU self-calibration method for equipped with a low-cost sensor. The goal our is to concurrently perform 3D pose estimation calibration based on optimization methods in unknown...

This article reports a new theory in physics: the of Observational Relativity (OR). The OR has discovered that all spacetime models and theoretical systems human physics must be branded with observation. uncovered root essence relativistic effects matter motion interactions presented spacetime: All are observational apparent phenomena. speed light is not really invariant; Spacetime curved. Galilean transformation Newtonian mechanics product idealized observation agent OA, presenting us...

Monocular depth estimation relied on RGB images is an important ill posed problem in ithe system of computer vision. Recently, people use the method deep learning to discuss this Most existing monocular algorithms convolution neural network. Depth based 2D has applications image segmentation, 3D object detection, robot navigation, tracking and autonomous driving. This paper gives a brief overview problem, reviews, evaluates discusses learning, looks forward direction further research face...

This paper studies the balance control of two-wheeled self-balancing robot on slope. The dynamic model was first established using Lagrange method, which proved correct by simulation results its zero-state response and zero-input response. A set equations were then obtained from kinetic model, Lyapunov method carried out to estimate stability controllability. Finally, robot's linear in equilibrium position obtained. quadratic optimal regulator designed attitude speed. show that slope...

To deal with the problem of stable walking quadruped robots on slopes, a gait planning algorithm framework for facing unknown slopes is proposed. We estimated terrain slope by attitude information measured inertial measurement unit (IMU) without relying robot vision. The crawl was adopted, and center gravity trajectory carried out based stability criterion zero-moment point (ZMP). Then, augmented random search (ARS) used to modulate parameters Bezier curve realize foot trajectory....

A two-loop cascade adaptive controller is proposed for a non-stable, non-linear, strong coupling system using backstepping and fuzzy neural network. The approach uses networks to approximate unknown nonlinear function. Then, use design realize self-balancing control of robot. simulation results indicate that both speediness stability performance in robot improved, there well trouble.

This paper focuses on a system to fulfill writer identification, which identifies person with his/her handwriting. A text independent method is proposed in this paper, involves no local feature analysis. First, preprocessing employed normalize script images. Then, 2D Gabor wavelet technique developed extract the global features. Finally K-nearest neighbor (K-NN) classifier designed identify identified Illustrative experiments are made 110 specimens of 50 people, and result 97.6% accuracy...

This paper proposed a hybrid approach of genetic algorithm (GA) and ant colony optimization (ACO) for the traveling salesman problem. In this approach, every chromosome GA is at same time an ACO. Whenever performs operation crossover mutation, firstly computes linkage strength between gene codes parental chromosome(s) according to pheromone matrix ACO, it then selects or mutation point(s) strength. A threshold generated classify as strong weak, segments parents are retained offspring far...

Aiming at the propose of preventing various barriers, which may cause collisions, during moving process a two-wheeled self-balanced robot, ultrasonic ranging theory and DSP TMS320F28335 are adopted as core controller by taking robot Hominid 3 platform. An obstacle avoidance control system for based on wave sensors is presented in this paper. employed to store environmental information, collected sensors. Particularly, fuzzy controltheory was introduced algorithm evaluation design controller....

This paper aims at a dynamic gesture recognition method based on the improved DTW algorithm. Firstly, 3D position of human skeletal points is obtained through analysis depth information, which by SDK Kinect sensor. 8 are selected as hand movement characteristics, and mathematical model established weighted distance. Then algorithm distortion threshold path constraints, used for training templates recognition. The results show that this can realize have good real-time performance robustness....

In this paper we study the dynamic modeling of a unicycle robot composed wheel, frame and disk. The can reach longitudinal stability by appropriate control to wheel lateral adjusting torque imposed is derived Euler-Lagrange method. controllability system are analyzed according mathematic model. Independent simulation using MATLAB ODE methods then proposed respectively. Through simulation, confirm validity two obtained models system, provide experimental platforms for designing balance controller.

This paper present a novel method to control the balance of two-wheeled robot by using reinforcement learning and fuzzy neural networks(FNN) which can guarantees convergence rapidity when model is not available agent has no prior knowledge. Furthermore it effectively task continuous states actions. The simulation experiment results demonstrate that only learn system in short time, but also maintain parameters change lot.

A method based on direct adaptive fuzzy control was proposed according to upstanding-balancing problem for Two-Wheeled Upstanding Robot. Different from the control, it didn't need design rules in beginning. And there not strict limit constant before input. Experiments showed that could be positive and negative, a function an interval with same sign. The result almost same. simulation had better of upstanding balancing robot even though initial angle larger. In contrast traditional scheme...

Inverted pendulum is a control system, with the feature of high order, muti-variable, non-linearity and unstable naturally. It quite important for us to study its balance stability in engineering field. In this paper, flywheel inverted pendulum, as an object be controlled, dynamic model has been established, mathematical which established certificated did system performance analysis. Then, linear model, fuzzy controller designed based on packet control. The method can decrease number rules...

For the problem of extracting feature steady-state visual evoked potential (SSVEP)-based brain-computer interface (BCI) system efficiently, a method based on independent component analysis (ICA) and Hilbert-Huang transform (HHT) is proposed in this paper. Firstly, Band-pass filter applied to preprocess electroencephalograph (EEG) SSVEP. Secondly, components are acquired from filtered signals with ICA. Thirdly, HHT decompose obtain intrinsic mode function (IMF) needed. Finally, frequency...

In the crowd navigation, reinforcement learning based on graph neural network is a promising method, which effectively solves poor navigation effect social interaction model and freezing behavior of robot in extreme cases. However, since information correlation human trajectory has not been involved its performance still needs improvement. Therefore, we proposed deep Social Spatial–Temporal Graph Convolution Network (SSTGCN) to handle problem, spatial–temporal taken advantage predict...

Associative learning, including classical conditioning and operant conditioning, is regarded as the most fundamental type of learning for animals human beings. Many models have been proposed surrounding or conditioning. However, a unified integrated model to explain two types much less studied. Here, based on neuromodulated synaptic plasticity presented. The bioinspired multistored memory module simulated VTA dopaminergic neurons produce reward signal. weights are modified according signal,...

This paper proposes a dual-loop adaptive decoupling control method based on single neuron PID controller (DADC-SNPID), balancing the wheeled robot. A unique mechanical and hardware structure of robot is designed its simplified mathematical model established using Newton-Euler equations according to actual parameters new balance motion proposed as Results simulation physical experiments are conducted illustrate effectiveness system under certain conditions.