Visual self-localization for nonholonomic mobile robots using a Hybrid Skip-list inspired Search Algorithm with a Gradient Policy

This paper presents a novel approach of using visual odometry for the self-localization of a maneuverable mobile robot configuration. This system is predominantly adapted to mobile robots that require autonomous positioning information at a low computation cost for real time applications. This system uses an over-head camera that captures images of the maneuvering mobile robot with two non-natural marker tags that tracks the vehicle and returns the vehicles orientation and its relative positional coordinates. This paper presents a brief discussion on typical procedures used to achieve self-localization of nonholonomic mobile autonomous robots. The developed Hybrid Skip-list Deterministic Search Algorithm with a Gradient Policy self-localization system is presented as it is integrated to a two wheel autonomous maneuverable mobile robot configuration. Finally, the theoretical and experimental results are presented that allow characterization of the performance of the self-localization system, illustrating the robustness and resilience of using a hybrid solution over a computation-hungry systematic solution.