This paper introduces the Clifford support vector machines (CSVM) as a generalization of real and complex-valued using geometric algebra. In this framework, we handle design kernels involving or product. approach, one redefines optimization variables multivectors. allows us to have multivector output. Therefore, can represent multiple classes according dimension algebra in which work. We show that apply CSVM for classification regression also build recurrent CSVM. The is an attractive...

The solution of the inverse kinematics mobile manipulators is a fundamental capability to solve problems such as path planning, visual-guided motion, object grasping, and so on. In this article, we present metaheuristic approach kinematic problem manipulators. approach, represent robot using Denavit–Hartenberg model. algorithm able taking into account platform. proposed avoid singularities configurations, since it does not require inversion Jacobian matrix. Those are two main drawbacks...

In this paper, an adaptive single neuron Proportional–Integral–Derivative (PID) controller based on the extended Kalman filter (EKF) training algorithm is proposed. The use of EKF allows online with faster learning and convergence speeds than backpropagation method. Moreover, propose PID approach includes a back-calculation anti-windup scheme to deal windup effects, which common problem in controllers. performance proposed shown by presenting both simulation experimental tests, giving...

Due to the complexity of manipulator robots, trajectory tracking task is very challenging. Most current algorithms depend on robot structure or its number degrees freedom (DOF). Furthermore, most popular methods use a Jacobian matrix that suffers from singularities. In this work, authors propose general method solve manipulators using metaheuristic optimization methods. The proposed can be used find best joint configuration minimize end-effector position and orientation in 3D, for robots...

In recent years, unmanned aerial vehicles (UAVs) have gained significant attention. However, we face two major drawbacks when working with UAVs: high nonlinearities and unknown position in 3D space since it is not provided on-board sensors that can measure its respect to a global coordinate system. this paper, present real-time implementation of servo control, integrating vision sensors, neural proportional integral derivative (PID), order develop an hexarotor image based visual control...

An approach to plan smooth paths for mobile robots using a Radial Basis Function (RBF) neural network trained with Particle Swarm Optimization (PSO) was presented in [1]. Taking the previous as an starting point, this paper it is shown that possible construct simple global path and then modify locally PSO-RBF, Ferguson splines or Bézier curves PSO, order describe more complex partially known environments. Experimental results show our fast effective deal

Nowadays, there are several meta-heuristics algorithms which offer solutions for multi-variate optimization problems. These use a population of candidate explore the search space, where leadership plays big role in exploration-exploitation equilibrium. In this work, we propose to Germinal Center Optimization algorithm (GCO) implements temporal through modeling non-uniform competitive-based distribution particle selection. GCO is used find an optimal set parameters neural inverse control...

The inverse kinematics of robotic manipulators consists finding a joint configuration to reach desired end-effector pose. Since is complex non-linear problem with redundant solutions, sophisticated optimization techniques are often required solve this problem; possible solution can be found in metaheuristic algorithms. In work, modified version the firefly algorithm for multimodal proposed kinematics. This provide multiple configurations leading same pose, improving classic performance....

SUMMARY An inverse optimal neural controller for discrete‐time unknown nonlinear systems, in the presence of external disturbances and parameter uncertainties, is presented. It based on a recurrent high‐order network trained with an extended Kalman filter‐based algorithm. The applicability proposed approach first tested via simulations electrically driven nonholonomic mobile robot, finally, methodology implemented real time. Copyright © 2012 John Wiley & Sons, Ltd.

The control of Unmanned Aerial Vehicles (UAV) is an important task in mobile robotics because its extensive use and the advantages over other kind robots like ground vehicles. In this paper, a PID controller for quadrotor based on Artificial Neural Network (ANN) proposed. neural network can handle limitations when controlling complex non linear systems. consists single neuron with three inputs their weights represent gains (K <sub xmlns:mml="http://www.w3.org/1998/Math/MathML"...

This article presents an approach to solve the inverse kinematics of cooperative mobile manipulators for coordinate manipulation tasks. A self-adaptive differential evolution algorithm is used as a global constrained optimization problem. model system proposed, considering with two omnidirectional platform n DOF. An objective function formulated based on forward equations. Consequently, proposed does not suffer from singularities because it require inversion any Jacobian matrix. The design...

Summary This work presents a neural observer‐based controller for uncertain nonlinear discrete‐time systems with unknown time‐delays. The proposed observer does not need previous knowledge of the model about system under consideration, neither value its parameters, delays, nor their explicit estimations. is based on network composed two recurrent high order networks (RHONNs) nonmeasurable state variables, one in parallel configuration, and measurable variables series‐parallel configuration....

With the increasing power of computers, amount data that can be processed in small periods time has grown exponentially, as importance classifying large-scale efficiently. Support vector machines have shown good results large amounts high-dimensional data, such generated by protein structure prediction, spam recognition, medical diagnosis, optical character recognition and text classiffication, etc. Most state art approaches for learning use traditional optimization methods, quadratic...

First-order algorithms have long dominated the training of deep neural networks, excelling in tasks like image classification and natural language processing. Now there is a compelling opportunity to explore alternatives that could outperform current state-of-the-art results. From estimation theory, Extended Kalman Filter (EKF) arose as viable alternative has shown advantages over backpropagation methods. Current computational advances offer review derived from EKF, almost excluded...

In this paper, we present a method for estimating GPS coordinates from visual information captured by monocular camera mounted on fixed-wing tactical Unmanned Aerial Vehicle at high altitudes (up to 3000 m) in GPS-denied zones. The main challenge odometry using aerial images is the computation of scale due irregularities elevation terrain. That is, it not possible accurately convert pixels image meters space, and error accumulates. contribution work reduction accumulated comparing with...