This work presents a robotic system that can navigate human environments, respond to speech commands, and perform simple tasks. To achieve this, a ballbot-type robot that balances and navigates on a single spherical wheel is used. This system possesses many advantages such as omnidirectional motion and agile mobility, which makes it apt to function in human environments.
In order to achieve smooth and agile navigation, prior work that formulates the system as 'differentially flat' is used. This formulation provides necessary lean angle trajectories used by the robot to move from one point to another. Leveraging this, a navigation framework using the concept of 'Velocity-Obstacles' is proposed to dynamically avoid human obstacles. An efficient and lightweight human tracking framework using 'SSD-Mobilenets' and 'Simple Online Realtime Tracking (SORT)' is also developed to facilitate efficient obstacle avoidance. Finally, speech capability is integrated into the system using Amazon-Alexa's extensive skills and capabilities.
To demonstrate various components of the system, two experiments are performed. In the first, the robot actively guides a user to necessary product locations in a mock grocery store. In the second, the robot performs 'Go, Look, and Tell tasks' in a household environment while avoiding moving human obstacles. Specifically, the robot goes to a particular location in the house, detects items, and reports them back to the user.
Ralph Hollis (advisor)
Zoom Participation. See announcement.