Autonomous Operation and Human-Robot Interaction on an Indoor Mobile Robot

Abstract

<p>MARVIN (Mobile Autonomous Robotic Vehicle for Indoor Navigation) was once the flagship of Victoria University’s mobile robotic fleet. However, over the years MARVIN has become obsolete. This thesis continues the the redevelopment of MARVIN, transforming it into a fully autonomous research platform for human-robot interaction (HRI). MARVIN utilises a Segway RMP, a self balancing mobility platform. This provides agile locomotion, but increases sensor processing complexity due to its dynamic pitch. MARVIN’s existing sensing systems (including a laser rangefinder and ultrasonic sensors) are augmented with tactile sensors and a Microsoft Kinect v2 RGB-D camera for 3D sensing. This allows the detection of the obstacles often found in MARVIN’s unmodified office-like operating environment. These sensors are processed using novel techniques to account for the Segway’s dynamic pitch. A newly developed navigation stack takes the processed sensor data to facilitate localisation, obstacle detection and motion planning. MARVIN’s inherited humanoid robotic torso is augmented with a touch screen and voice interface, enabling HRI. MARVIN’s HRI capabilities are demonstrated by implementing it as a robotic guide. This implementation is evaluated through a usability study and found to be successful. Through evaluations of MARVIN’s locomotion, sensing, localisation and motion planning systems, in addition to the usability study, MARVIN is found to be capable of both autonomous navigation and engaging HRI. These developed features open a diverse range of research directions and HRI tasks that MARVIN can be used to explore.</p>