Adaptive Parameter EXploration (APEX): Adaptation of robot autonomy from human participation

Abstract

The problem of Adaptation from Participation (AfP) aims to improve the efficiency of a human-robot team by adapting a robot's autonomous systems and behaviors based on command-level input from a human supervisor. As a solution to AfP, the Adaptive Parameter EXploration (APEX) algorithm continuously explores the space of all possible parameter configurations for the robot's autonomous system in an online and anytime manner. Guided by information deduced from the human's latest intervening commands, APEX is capable of adapting an arbitrary robot system to dynamic changes in task objectives and conditions during a session. We explore this framework within visual navigation contexts where the human- robot team is tasked with covering or patrolling over multiple terrain boundaries such as coastlines and roads. We present empirical evaluations of two separate APEX-enabled systems: the first, deployed on an aerial robot within a controlled environment, and the second, on a wheeled robot operating within a challenging university campus setting. I. INTRODUCTION We define Adaptation from Participation (AfP) for a human-robot team as the problem of dynamically adjusting configuration parameters of an autonomous robot system (henceforth referred to as a robot autonomy) with the aim of maximizing the team's overall efficiency in terms of improved task performance and reduced human workload. We propose an online and anytime solution to AfP called Adaptive Parameter EXploration (APEX), which uses mul- tiple competing parameter hypotheses that we refer to as particles to simultaneously explore the parameter space of the robot autonomy. Particles optimize their hypotheses using information from the human's latest intervening commands to search for configuration settings that can effectively handle the evolving task objectives and environmental conditions that occur during a session. We evaluate two APEX-enabled autonomies for terrestrial and aerial visual navigation tasks wherein human operators and autonomous robots collaborate to sequentially cover or patrol through different terrain boundaries such as shorelines and roadsides. Human-robot teams have the potential to solve very chal- lenging tasks as they combine the heightened dexterity and comprehensive planning capabilities of autonomous robots with the keen instincts and creative problem solving skills of humans. A key concern, however, is that it is often difficult for the operator to configure the robot autonomy by hand, e.g. when tuning controller gains, adjusting learning rates, etc. Such manual adjustments are especially challenging to perform during an active task session and also for systems