ROS Framework for Perception and Dual-Arm Manipulation in Unstructured Environments

Abstract

In a near future, robotic systems are expected to be able to confront more complex tasks in challenging scenarios. In this context, intelligent perception and dual-arm robotic manipulation capabilities are crucial for improving the autonomy and reliability of these systems. This paper addresses the development of an experimental platform conceived to facilitate the design and assessment of new perception and dual-arm control algorithms in unstructured environments. The proposed testbed is composed of a dual-arm robotic configuration endowed with a visual perception system and a simulation and control platform implemented in ROS (Robot Operating System). The robotic configuration consists of two manipulator arms of 6-DOF (Kinova MICO™) with brushless DC actuators controlled directly through PID controllers, whereas the perception system is formed by a high resolution RGB camera and a Time-of-Flight camera. ROS provides an open source collection of software frameworks, which simplify the task of creating complex and robust robot behaviours across a wide variety of robotic systems. The proposed approach will enable the easy testing and debugging of new applications with zero-risk damage to the real equipment.