A subsumption architecture for autonomous docking in low-order complexity environments

This thesis presents a subsumption architecture for autonomous vision-based docking by a four-wheeled agent with ultrasonic obstacle detection. Docking is performed with a stationary docking receptacle. The effectiveness of the docking methodology is empirically studied with a four-wheeled autonomous agent in environments of low-order complexity as defined by metrics generated in this work.; Docking methodologies for autonomous agents are important to enable refueling fuel stations. Despite the increasing complexity of the environments in which agents interact, previous research in docking has presupposed the environment to be well known, and there has not been a metric that has defined the complexity of the studied environments as a consequence of obstacles and orientation of the agent with respect to the docking receptacle. The current research shows that the generated subsumption architecture can successfully dock head-on with no obstacles present over 90% of the time on a four-wheeled autonomous agent. It is also demonstrated that vision-based docking failures in this model are chiefly a result of the sharpness of the angle of approach a four-wheeled agent takes toward the docking receptacle, and not directly correlated to the number of obstacles in the environment.