Integracja czujników inercyjnych z konstrukcją robota humanoidalnego cz. II

W niniejszym artykule przedstawiono algorytmy sterowania robotem humanoidalnym firmy Futaba. Zaprezentowano wykorzystanie sztucznych sieci neuronowych, jako alternatywnego sposobu obliczenia kinematyki odwrotnej oraz wykorzystanie środowiska Microsoft Robotics Developer Studio do tworzenia złożonych, wielowątkowych aplikacji szeroko stosowanych w robotyce. Ponadto pokazano zastosowanie środowiska symulacyjnego VSE (Visual Simulation Environment) w procesie prototypowania algorytmów sterujących.

This paper presents a control system for a Futaba humanoid robot [1]. It is equipped with a controller board based on a microcontroller with an ARM7TDMI core, a full set of inertial sensors and a 2.4 GHz wireless communication module. The controller uses the wireless communication module to send information about the robot state to a PC on which the controlling application is run. This paper focuses on the software part of the presented system (the hardware part has been presented in the first part of this paper). In order to develop a controlling algorithm, an analysis of robot kinematics was made and equations for direct kinematics were derived in consistency with the Denavit-Hartenberg convention. To eliminate the necessity of designating equations for inverse kinematics, which can be very complex due to the kinematic redundancy of the robot, an artificial neural network was used (Fig. 2). The application was written using the Microsoft Robotics Developer Studio designed for creating complex, multithread, distributed and scalable applications used in robotics. The application uses the data acquired by radio to implement the walking and balance-keeping algorithms. For visualization of the robot movement, testing and development of the algorithms without the risk of damaging the robot, a simulation in the Visual Simulation Environment, a part of the Microsoft Robotics Developer Studio, was created. A 3D model of the robot was used in this simulation (Fig. 4).