Perception-Aware Human-Assisted Navigation of Mobile Robots on Persistent Trajectories

Abstract

We propose a novel shared control and active perception framework combining the skills of a human operator in accomplishing complex tasks with the capabilities of a mobile robot in autonomously maximizing the information acquired by the onboard sensors for improving its state estimation. The human operator modifies at runtime some suitable properties of a persistent cyclic path followed by the robot so as to achieve the given task (e.g., explore an environment). At the same time, the path is concurrently adjusted by the robot with the aim of maximizing the collected information. This combined behavior enables the human operator to control the high-level task of the robot while the latter autonomously improves its state estimation. The user's commands are included in a task priority framework together with other relevant constraints, while the quality of the acquired information is measured by the Shatten norm of the Constructibility Gramian. The user is also provided with guidance feedback pointing in the direction that would maximize this information metric. We evaluated the proposed approach in two human subject studies, testing the effectiveness of including the Constructibility Gramian into the task priority framework as well as the viability of providing either visual or haptic feedback to convey this information metric.

Autonomy in Physical Human-Robot Interaction: A Brief Survey

“…The latter approach, allows the user to interact in real-time with a remote inspection robot at the trajectory level: the user can steer the reference path of an autonomous mobile robot (e.g., a UAV) by acting on path parameters that are simultaneously affected by an autonomous algorithm to ensure collision avoidance, path regularity, proximity to points of interest [41] or maximize the collected environmental information [42].…”

Informing Real-Time Corrections in Corrective Shared Autonomy Through Expert Demonstrations

“…The approach presented in this paper focuses on when the corrections are generalized beyond kinematic variables and inputted using an external controller. Existing methods for corrections using an external controller [11], [12], [13] allow for path transformations and policy learning of UAVs and mobile robots. However, similar to the methods in pHRI, the main emphasis is on corrections to kinematics and on updating the trajectory or policy rather than using task knowledge to inform possible corrective input.…”

Online Decentralized Perception-Aware Path Planning for Multi-Robot Systems

“…Among the many possible measures for quantifying the amount of information about the state of a nonlinear system, in this work we consider the so-called Gramians [32], [33] which have already been successfully applied to optimal estimation problems for single mobile robots [25], [34], [35]. Gramians indeed represent a general observability measure not related to the particular filter used for state estimation (e.g., a EKF).…”