Autonomous Operations of Mobile Robots in a Full Range of Environments

Abstract

Autonomous Operations of Mobile Robots in a Full Range of Environments

Xiaoping Yun, James Calusdian, and Matthew Audette

Abstract: The objective of this project is to develop autonomous capabilities of mobile robots in a full range of real-world indoor and outdoor environments. The entire NPS campus including all the base ground and building structures are used as a testbed for testing and evaluation. The ultimate goal is to develop autonomous capabilities that allow mobile robots to travel from any one location on the campus to another location. Examples include going from Building 436 (Police Service) to the front entrance of Dudley Knox Library, or from Spanagel-429 to Hermann Hall Barbara McNitt Ballroom. To make this possible, the robots are required to navigate through a wide range of indoor and outdoor environments. A fleet of P3-AT wheeled mobile robots ruggedized for rough-terrain environments are utilized in the project. Through this large-scale experimentation, it is expected to learn lessons and discover challenges in deploying autonomous robots in complex, real-world situations. It is hoped that the knowledge gained on navigating the NPS campus is applicable to other installations and can be used to support battlefields in urban environments.

This project is leveraged on the prior efforts in developing robot navigation and mapping algorithms for indoor and outdoor environments. The early efforts were limited to navigate in a laboratory space and a portion of the outdoor Academic Quad area between Root Hall and Bullard Hall. This project seeks to expand the study to cover the entire NPS campus. The robots are expected to navigate from one building to another, approach a building entrance using handicap wheelchair ramps if necessary, travel from one floor to another via elevators and from any room to another inside a building. The research is carried out by multiple thesis students in stages, and is focused on obstacle avoidance, localization, mapping, path planning, sensor integration, and data fusion. Navigation algorithms rely on real-time sensor measurements of the environment as well as the pre-existing map or building data. Google map, the campus map data, and the building floor layout data from the Public Works Office are used to build a pre-existing map to be used by the navigation algorithms. A combination of sensor suite including LiDAR, sonar, infrared, CCD camera, IMU sensor, Kinect sensor, and GPS receiver is integrated for sensing the environment in real time to identify objects that are not registered in the map data. The pre-existing map is updated whenever a new stationary object is detected by the onboard sensors.