EAGER: Modeling Intent Communication Pathways for Human-Autonomous System Collaboration

Duke University, Durham NC

Investigators

Abstract

Autonomous systems show great promise for enhancing safety, however previous research in human-automation interaction has also demonstrated that adding automation improperly may actually result in highly undesirable dynamic interactions. Furthermore, automating a task within a larger system may transfer the operator's workload from one physical or cognitive resource to another, thereby merely changing the task, rather than improving it. Thus it is crucial when designing new technologies to consider carefully how the operator will interact with system automation, particularly with autonomous systems that are based on stochastic, as opposed to deterministic, reasoning. This EArly-concept Grant for Exploratory Research (EAGER) project will determine how to design safe autonomous systems that have awareness of the intent of humans in and around the system, with reciprocal relationships for those same humans. Previous work in human-centered design focused on interactions between an engineered system and a human operator. This project adds consideration of interfaces designed to communicate with nearby humans, with an emphasis on designing autonomous systems that consider the intent of both the external individual and internal operator. This objectives of this project are as follows: (1) perform a comprehensive analysis of human-system interactions to identify methods for representing intent in interactions between an engineered system and individuals within and outside the system; (2) develop models for computer interpretation of human intent, including the intent of the system operator and exogenous actors; and (3) identify methods to communicate system intent to the operator and to exogenous actors. The resulting models will be verified by adapting the model to the driverless car domain -- an area where technology has outpaced understanding of how humans integrate into the system. The project will also extend the modeling approach to the manufacturing domain using a pick-and-place Baxter robot. In both settings, the researchers will study how elements of the environment, the computation systems of the robots, and unique traits of humans can be modeled to represent intent communication pathways that need to be instantiated in the system or in the world around the system. The effort will also study the needs of stakeholders including endogenous and exogenous actors (e.g., pedestrians and workers in a plant near a robotic forklift). Applying the models across multiple domains will determine the extent to which the models of both entities can be generalized.