Autonomous Robots in Unknown Environments: Online Semantic SLAM and Preference-based Planning
Abstract: This dissertation aims to enable the robot to perform a set of tasks with a preference order in a priori unknown environment, a fundamental capability for service robots operating in domestic environments, and a research problem yet to be addressed by the current literature. This dissertation leverages our original research in semantic SLAM, object search, and preference-guided planning to build a robot system that can address this overarching research problem.
The contributions of this dissertation can be summarized as follows: First, a custom semantic SLAM algorithm is proposed. The object-level data association problem is addressed, and a map object initialization scheme with a higher success rate is proposed. A novel semantic loop closure method is also designed utilizing the object covisibility graph maintained in the map database of semantic SLAM. This loop closure method can distinguish similar scenes and avoid false loop closure. Secondly, the object search problem is addressed, which can be considered a simplified version and a milestone of the original research problem. Semantic prior knowledge is encoded into a Bayesian Network to facilitate the search for the target object. Thirdly, for preference-guided planning, the dilemma between exploitation and exploration is solved by formally appending the exploration task to the original task set with a preference order. Theoretic guarantees for the robot's behavior are proved mathematically. Finally, the synthesized policy for robot motion is updated online to adapt to the expanding knowledge of the world, with its time determined by the divergence of a proven upper bound score.
Extensive experiments in both synthetic and real-world environments have demonstrated the effectiveness and efficiency of individual components in the perception and planning stacks, as well as the overall capability of the proposed system to perform a set of tasks with a preference order in a priori unknown environment.
Advisors: Professor Jing Xiao and Professor Jie Fu
Committee: Professor Carlo Pinciroli, Professor Nitin Sanket, and Professor Xinming Huang
Zoom Link: https://wpi.zoom.us/j/6641279149