Real-time low-level software framework for controlling robot devices

Abstract

In this dissertation, a real-time robot software framework for controlling low-level robot devices named PODO is presented. To meet the demands of many users in the different robotic fields, PODO was developed with four philosophical goals: oriented to control and real-time, designed for collaborative development, optimized to embedded system, and expandable for both hardware and software. To achieve such goals, PODO was designed to have a set of multiple processes. It was also structured on the framework consisting of five hierarchical layers: End Devices, Device Process (Daemon), Shared Memory, User Processes (ALs), and External Processes. For a hard real-time with a bounded action delay, Daemon provides a real-time thread and controls End Devices

ALs are synchronized with Daemon through Shared Memory. The Shared Memory possesses the data mapping hardware configurations, which enclose details of controlling the low-level robot devices so the users do not need to understand a way how Daemon controls the robot devices. The ALs can be used by multiple users in order to generate the motion references, and External Processes can be used as a medium of other robot software such as OPRoS and ROS in order to be expandable. Actually, in this dissertation,ROS and PODO are integrated, and SLAM and Navigation are demonstrated as an example. It also shows that Gazebo, a dynamics simulator, replaces the robot platform, and the PODO-ROS-Gazebo system is introduced. The real-time performance of PODO was verified by means of acquiring an accurate clock time of the system. The experiment shows that the time for synchronization is bounded within a certain time range. Other features in PODO were also verified through many practical applications. The results show that PODO can be a good candidate for a standard as the robot software platform for controlling the robot devices.