Control and Perception of Robotic Movement with Styles

Robots are usually thought of as tools that carry out different tasks with functional movements. Robots that need to operate in human-facing environments will require complex options for modifying their movements to communicate changing state information. However, current robotic platforms are limited in reproducing the variation of movement in human perception, such as Baxter Research Robot from Rethink Robotics and Nao from Aldebaran Robotics. This research focuses on improving the feature variation of robotic movement to make robots more accessible, engaging, and collaborative when interacting with humans.

In this research, approaches are designed for solving the problem that occurs when the generated trajectories in high-level controller exceed the physical limits of a particular robotic platform. This work aims to guarantee the trajectories generated by prior high-level controllers are executable on physical robotic platforms. Two of the approaches are used as part of a web-based application that leverages ROS, MATLAB, and onboard low-level controllers to show how the methods can be applied with the technical details abstracted away for a user. This system is implemented on Rethink Robotics Baxter Research Robot with different selections of quality parameters to demonstrate the methods.

To make the human-robot interaction more intuitive and more effective for intent recognition, perceptually meaningful sound is supplemented to the robotic movement. Human vocalization responses to the videos of different simulated robotic movements were recorded and these recordings were analyzed to study how sonic features map to features of movement in human perception. The mapping of features of sound and movement enables us to create appropriate sounds to accompany robotic movement to help convey movement qualities and make it more expressive. To further improve the variations in robotic movement, a new method for generating the movement reference trajectories with exaggerated variations is proposed. This method is designed based on the affinities between Effort and Space in Laban/Bartenieff Movement System (LBMS).

A user study is carried out for testing whether accompanying movement with sound and the method of generating movement trajectories with spatial affinities are helpful in improving the expressivity of robotic movement. This user study consists of surveys of perceiving the qualities in robotic movement in stick figure animations and videos of Baxter Research Robot. The user study data was analyzed quantitatively using statistical tests.

The existing perceived variations in robotic movement were increased by generating physically feasible movement trajectories for a robot to carry out, accompanying the movements with appropriate sounds and generating more various reference trajectories for a robot to track.