Sensor-aligned coverage planning for mobile robots in complex 3D environments

Efficient and effective monitoring of radiation levels is crucial for the long-term maintenance of nuclear facilities and the safety of nuclear workers. Unlike gamma and beta radiation, which are detectable some distance from the radiation source, alpha radiation detection requires that the source be less than a centimeter from the sensor to provide consistent readings. Despite its small range, alpha radiation is deadly if ingested or inhaled making accurate surveys critical for worker safety. Detection is typically performed manually by dedicated workers, a long and inefficient process given the sensing requirements. Mobile robotics provides a promising alternative to this process, capable of providing services which augment and accelerate the process alongside human workers.

This thesis presents a robot agnostic framework for a wheeled mobile robot equipped with a large alpha detector to autonomously generate complete coverage plans of a facility floor, the largest surface of interest in routine surveys. In the generation of this plan, it accounts for the full 3D geometry of the robot and the environment by subdividing the robot model into cylindrical portions which are compared against the incoming sensor data. This collision detection method is used to subdivide each grid cell in the workspace into a set of valid poses centered around the radiation sensor. These are then connected to adjacent poses to form a graph in which each node is a valid robot state. By applying existing coverage planning algorithms to this graph, complete coverage of the workspace is achieved. Tests showed successful planning in simulated and real life test environments on two different robot models, indicating that this model is applicable to the alpha radiation survey problem.