Adaptive Autonomy Research Articles

Human factors considerations for the context-aware design of adaptive autonomous teammates

Despite the gains in performance that AI can bring to human-AI teams, they also present them with new challenges, such as the decline in human ability to respond to AI failures as the AI becomes more autonomous. This challenge is particularly dangerous in human-AI teams, where the AI holds a unique role in the team’s success. Thus, it is imperative that researchers find solutions for designing AI team-mates that consider their human team-mates’ needs in their adaptation logic. This study explores adaptive autonomy as a solution to overcoming these challenges. We conducted twelve contextual inquiries with professionals in two teaming contexts in order to understand how human teammate perceptions can be used to determine optimal autonomy levels for AI team-mates. The results of this study will enable the human factors community to develop AI team-mates that can enhance their team’s performance while avoiding the potentially devastating impacts of their failures.

Autonomous behaviour and the limits of human volition

Humans and some other animals can autonomously generate action choices that contribute to solving complex problems. However, experimental investigations of the cognitive bases of human autonomy are challenging, because experimental paradigms typically constrain behaviour using controlled contexts, and elicit behaviour by external triggers. In contrast, autonomy and freedom imply unconstrained behaviour initiated by endogenous triggers. Here we propose a new theoretical construct of adaptive autonomy, meaning the capacity to make behavioural choices that are free from constraints of both immediate external triggers and of routine response patterns, but nevertheless show appropriate coordination with the environment. Participants (N = 152) played a competitive game in which they had to choose the right time to act, in the face of an opponent who punished (in separate blocks) either choice biases (such as always responding early), sequential patterns of action timing across trials (such as early, late, early, late…), or predictable action-outcome dependence (such as win-stay, lose-shift). Adaptive autonomy was quantified as the ability to maintain performance when each of these influences on action selection was punished. We found that participants could become free from habitual choices regarding when to act and could also become free from sequential action patterns. However, they were not able to free themselves from influences of action-outcome dependence, even when these resulted in poor performance. These results point to a new concept of autonomous behaviour as flexible adaptation of voluntary action choices in a way that avoids stereotypy. In a sequential analysis, we also demonstrated that participants increased their reliance on belief learning in which they attempt to understand the competitor's beliefs and intentions, when transition bias and reinforcement bias were punished. Taken together, our study points to a cognitive mechanism of adaptive autonomy in which competitive interactions with other agents could promote both social cognition and volition in the form of non-stereotyped action choices.

Uncrewed undersea vehicle embedded autonomous target recognition in synthetic aperture sonar imagery

The opacity of the ocean to electromagnetic signals makes communication with Uncrewed Undersea Vehicles (UUVs) extremely difficult, especially for situations where the vehicle is operating far from its center of command and control, and makes mid-operation data exfiltration for external processing untenable. This leaves operator intervention and control limited necessitating the need for advanced and adaptive autonomy that can create situational awareness from mission data provided by vehicle sensors. Long range high-resolution imagery generated by synthetic aperture sonar (SAS) is a high fidelity data product that can be further processed by both machine learning techniques and traditional image processing techniques to create computer aided detection and classification (CAD/CAC) capabilities to feed that developing autonomy need. This work examines two separate detection and classification pipelines that have been used in parallel to create contact sets that are actionable for onboard decision processes.

The lived, living, and behavioral sense of perception

With Jan Degenaar and Kevin O’Regan’s (D&O) critique of (what they call) ‘autopoietic enactivism’ as point of departure, this article seeks to revisit, refine, and develop phenomenology’s significance for the enactive view. Arguing that D&O’s ‘sensorimotor theory’ fails to do justice to perceptual meaning, the article unfolds by (1) connecting this meaning to the notion of enaction as a meaningful co-definition of perceiver and perceived, (2) recounting phenomenological reasons for conceiving of the perceiving subject as a living body, and (3) showing how the phenomenological perspective does a better job at fulfilling D&O’s requirement for grounding notions of mentality in ‘outer’ criteria than they do. The picture that thus emerges is one of perceptual meaning as an integration of lived, living, and behavioral aspects – a structure of behavior that cannot be captured by appeal to sensorimotor capacities alone but that is adequately illuminated by the enactive notion of adaptive autonomy.

Overcoming the Lumberjack Effect Through Adaptive Autonomy

Research shows that there are a variety of performance advantages to increasing the degrees of autonomy that human-supporting autonomous systems possess, including organizational productivity and the reduction of human workload. Yet, not all the consequences of increasing autonomy are positive. One such negative consequence is the risk of human operators becoming more incapable of responding to system failures as autonomy levels increase, a phenomenon known as the Lumberjack Effect. This paper proposes a conceptual model for using adaptive autonomy as a means to avoid the risks of this phenomenon while retaining the advantages of higher autonomy levels in earlier stages of an organization’s work cycle. We apply our model to two research-based scenarios and discuss future research necessary to validate the model. This model provides the Human Factors community with a possible solution to the debate over the risks of designing human-supporting autonomous systems with higher degrees of autonomy.

Characterization of Indicators for Adaptive Human-Swarm Teaming.

Swarm systems consist of large numbers of agents that collaborate autonomously. With an appropriate level of human control, swarm systems could be applied in a variety of contexts ranging from urban search and rescue situations to cyber defence. However, the successful deployment of the swarm in such applications is conditioned by the effective coupling between human and swarm. While adaptive autonomy promises to provide enhanced performance in human-machine interaction, distinct factors must be considered for its implementation within human-swarm interaction. This paper reviews the multidisciplinary literature on different aspects contributing to the facilitation of adaptive autonomy in human-swarm interaction. Specifically, five aspects that are necessary for an adaptive agent to operate properly are considered and discussed, including mission objectives, interaction, mission complexity, automation levels, and human states. We distill the corresponding indicators in each of the five aspects, and propose a framework, named MICAH (i.e., Mission-Interaction-Complexity-Automation-Human), which maps the primitive state indicators needed for adaptive human-swarm teaming.

Interactive Human–Robot Skill Transfer: A Review of Learning Methods and User Experience

Generalizing the operation of robots in dynamical environments regardless of the task complexity is one of the ultimate goals of robotics researchers. Learning from demonstration approaches supported by transfer learning and user feedback offer a remarkable solution to achieve generalization. The main idea behind such approaches is teaching robots new skills with human instructors and training parametric models with data from demonstrations to achieve and update the desired skills under changing conditions. Herein, the applications of skill transfer with reinforcement learning algorithms and the effect of user experience (UX) on learning from demonstration approaches are reviewed. This review outlines the importance of considering and evaluating UX during human–robot interaction and, especially, robot teaching. A detailed view on the relations between robot learning and UX is provided and approaches for future improvements are derived. Finally, adaptive autonomy sharing between the robot and the user during teaching is presented as a promising approach to enhance the interaction by exploiting user feedback. In the long run, interactive and user‐centered human–robot skill transfer is expected to reduce cognitive and physical load of the user. Discussion on future research questions aiming to improve learning process and semiautonomous behavior concludes the review.

Adaptive human-in-the-loop multi-target recognition improved by learning

Machine learning algorithms have been designed to address the challenge of multi-target recognition in dynamic and complex environments. However, sufficient high-quality samples are not always available for training an accurate multi-target recognition classifier. In this article, we propose a generic human-in-the-loop multi-target recognition framework that has four collaborative autonomy levels, and it allows adaptive autonomy level adjustment based on the recognition task complexity as well as the human operator’s performance. The human operator can intervene to relabel the collected data and guarantee the recognition accuracy when the trained classifier is not good enough. Meanwhile, the relabeled data are used for online learning which further improves the performance of the classifier. Experiments have been carried out to validate the proposed approach.

Method Variance in Adolescents', Mothers', and Observers' Reports of Peer Management: Nuisance or Information?

The purpose of this investigation was to examine discrepancies among three informants' (adolescents, mothers, and observers) reports of maternal consulting in regard to peer relationships and the relation of the discrepancies to four social adjustment variables (prosocial behavior, loneliness, positive friendship quality, and physical victimization). An ethnically diverse sample of 70 early adolescents (51 % female) and their mothers participated in this multimethod investigation. Adolescent reports of parental consulting, but not mother or observer reports, were significantly associated with adolescent reports of four psychosocial outcomes. Recognizing that comparison of equivalent regression models can inform interpretations of the data such as the monomethod associations found in this study, this article describes and discusses a strategy for analyzing data from multiple informants. The associations of adolescent reports of parental consulting with loneliness and physical victimization could be fully explained in terms of adolescent bias or other systematic variance uniquely associated with adolescent reports, but those with prosocial behavior and positive friendship quality could not. The view that discrepancies between mother and adolescent reports of parental consulting reflect poor relationship quality appeared most applicable in models of positive friendship quality, somewhat applicable in models of prosocial behavior and physical victimization, and not applicable in models of loneliness. The view that discrepancies might reflect normative and adaptive autonomy was not supported. In addition to adding to our understanding of maternal consulting in regard to peer relationships, it is also hoped that the analytic approach developed for this study will stimulate developments in research that uses multiple informants.

Environmentally adaptive autonomy for acoustic communication and navigation in the new Arctic

The particular sensitivity of the Arctic to climate change is well established. and the significance to undersea operations can be dramatic. As part of the recent ICEX16 US Navy Exercise in the Beaufort Sea, MIT deployed an autonomous underwater vehicle with a towed hydrophone array below the ice cover for assessing the climate induced changes to the undersea ambient noise environment. The safe underwater operation depended on navigation updates from the submarine tracking range being communicated to the vehicle for fusion with the onboard inertial navigation. However, the changes in the environment severely deteriorated the tracking performance compared to previous deployments. The reason was clearly associated with a previously observed neutrally buoyant layer of warm Pacific water persistently spreading throughout the Beaufort Sea, which severely alters the acoustic environment with dramatic effects for both long and short range acoustic sensing, communication and navigation. This paper describes the effects observed and discusses how robust acoustic connectivity in this environment makes it paramount that the manned or unmanned undersea platforms are capable of adapting to the environment for sensing, communication, and navigation. [Work supported by the Office of Naval Research.]

Autonomous and Adaptive Underwater Plume Detection and Tracking with AUVs: Concepts, Methods, and Available Technology

An autonomous underwater vehicle (AUV) equipped with environmental sensors and an on-board autonomy system can greatly increase the efficiency of environmental data collection and the synopticity of the data set collected simply by autonomously adapting its motion to changes it senses in its local environment. One application of this is tracking ocean features in an unknown ocean environment. This can be accomplished with one or multiple AUVs collaborating in near-real-time using acoustic communications. To further explore one example of this application, this paper focuses on using multiple AUVs to track underwater plumes. We evaluate various types of plumes (e.g., hydrothermal vent plumes, algal blooms, oil leaks), how each plume type may be detected and its spatial extent determined, what types of sensors can be used, and how AUVs can be employed to autonomously and adaptively track these dynamic plumes. Since AUVs vary significantly in design, mobility, deployment duration, on-board processing power, etc., it is also necessary to consider the best choice of AUV (or combination of AUVs) to track a plume. Thus, an operator/scientist's choice of AUV type(s) will likely depend the type of plume to be tracked, or vice versa. Since most underwater plumes are highly spatiotemporally dynamic, employing environmentally adaptive autonomy to track them with a fleet of AUVs is one of the most efficient ways to do so, given today's technology.

Impact of Two Adjustable-Autonomy Models on the Scalability of Single-Human/Multiple-Robot Teams for Exploration Missions

The aim of this study was to evaluate two models for adjusting autonomy in mobile robots to find out the best way for the operator to interact with the system with as many robots as possible. The first model is the most used in mobile robots; the second proposes a flexible autonomy management. There are different ways of adjusting the autonomy level in man-machine systems: adjustable autonomy, in which the operator has the initiative over the autonomy level; adaptive autonomy, in which the autonomy level is adjusted depending on the task and context; and mixed initiatives. One of the drawbacks of using adjustable autonomy is that it is claimed not to be flexible enough, resulting in a high operator workload. We propose and evaluate a flexible adjustable autonomy model for robot-team supervision. Two experiments were designed to test the scalability and performance of the man-machine system with two alternative configurations for the autonomy management. The independent variable is the number of robots, and the measured variable is the man-machine system performance. The experiments are between subjects. We have used ANOVA and Bonferroni post hoc analysis for analyzing the results. On the basis of these analyses,we conclude that a flexible adjustable autonomy model results in better performance than the classic, rigid one, in which the operator directly chooses the autonomy level. Flexible autonomy adjustment permits one operator to control a team of robots with better results in terms of performance and robot use, as he or she can directly act at the error level, leaving the responsibility of readjusting and resuming the task to the system and hence reducing the operator's workload. The results can be applied to exploration robotics, mainly, in which one operator controls a team of robots. In general, these principles can be extended to other single-man/multiple-machine systems.

Adaptive Preferences and Procedural Autonomy

Capabilities theorists hold that adaptive preference (APs) are problematically shaped by deprivation, and that they thus merit public interrogation. However, it is unclear what differentiates APs from preferences worthy of public respect. Thinking of APs as procedurally non‐autonomous promises grounds on which to distinguish them without compromising respect for moral pluralism. Using examples from gender and development practice, I argue that — despite the appeal of this route — there are deep problems with thinking of APs as non‐autonomous. Conceptions of APs as non‐autonomous do not identify APs in a way consistent with our intuitions and fail to provide appropriate practical guidance to public institutions interested in interrogating APs. I suggest in the conclusion that identifying APs requires a theory of the good.

An Expert System Realization of Adaptive Autonomy in Electric Utility Management Automation

Abstract: Earlier we introduced a novel framework for implementation of Adaptive Autonomy (AA). This study presents an expert system realization of the AA framework, referred to as Adaptive Autonomy Expert System (AAES). The proposed AAES is based on the extracted rules from the Expert’s Judgment on proper Levels of Automation (LOA) for various environmental conditions, modeled as Performance Shaping Factors (PSFs). Decision fusion method is used as expert system inference engine, where eight decision fusion methods are developed as prospective ones. The AAES is realized in the practical case of electric power Utility Management Automation (UMA) for the Greater Tehran Electricity Distribution Company (GTEDC). The practical list of PSFs and the judgments of GTEDC’s experts are used as the expert system rule base in this research. The results of implementing the proposed AAES to GTEDC’s network are evaluated according to two criteria: average error and error margin. Five out of eight decision fusion methods are proven to be suitable inference engines, due to both criteria. Evaluation of the results shows that the proposed AAES can estimate proper LOAs for GTEDC’s UMA system, which change due to the changes in PSFs; thus providing a dynamic (adaptive) LOA scheme for UMA.

A Framework for Implementation of Adaptive Autonomy for Intelligent Electronic Devices

A Framework for Implementation of Adaptive Autonomy for Intelligent Electronic Devices

DYNAMIC ADAPTIVE AUTONOMY IN MULTIAGENT SYSTEMS: REPRESENTATION AND JUSTIFICATION

Autonomy is an often cited but rarely agreed upon agent characteristic. Although no definition of agent autonomy is universally accepted, the concept of adaptive autonomy promises increasingly flexible and robust agent-based systems. In general, adaptive autonomy gives agents the ability to seek help for problems or take initiative when otherwise they would be constrained by their design to follow some fixed procedures or rules for interacting with other agents. In order to access these benefits, this article provides a core definition and representation of agent autonomy designed to support the implementation of adaptive agent autonomy. This definition identifies "decision-making control" governing the determination of agent goals and tasks as the key dimension of agent autonomy. In order to gain run-time flexibility and any associated performance improvements, agents must be able to dynamically adapt their autonomy during system operation. This article justifies the implementation of dynamic adaptive autonomy through a series of experiments showing that a multiagent system operating under dynamic adaptive autonomy performs better than a multiagent system operating under fixed autonomy for the same changing run-time conditions.

Toward flexible and fault tolerant intelligent manufacturing: Sensible agents in shop-floor control

The use of agent-based systems may offer significant advantages over centralized hierarchical manufacturing systems, including distribution of control/processing and data as well as adaptable automated or semiautomated problem solving. However, simply applying any agent-based paradigm to manufacturing domains may not be enough to address the various demands of these domains. Manufacturing domains require agent-based problem solving to be flexible and tolerant of faulty information (e.g., due to faulty sensors), equipments, and communication links. These issues are addressed by research on Sensible Agents, capable of 1) Dynamic Adaptive Autonomy and 2) explicit Perspective Modeling. Dynamic Adaptive Autonomy allows Sensible Agents to change problem-solving roles during system operation to address dynamic factory floor conditions, while Perspective Models internal to a respective Sensible Agent serve to maintain knowledge about the agent itself, other agents, and the environment. The application of the Sensible Agent paradigm to a shop-floor control is demonstrated.

Dynamic adaptive autonomy in multi-agent systems

Multi-agent systems require adaptability to perform effectively in complex and dynamic environments. This article shows that agents should be able to benefit from dynamically adapting their decision-making frameworks. A decision-making framework describes the set of multi-agent decision-making interactions exercised by members of an agent group in the course of pursuing a goal or set of goals. The decision-making interaction style an agent adopts with respect to other agents influences that agent's degree of autonomy. The article introduces the capability of Dynamic Adaptive Autonomy (DAA), which allows an agent to dynamically modify its autonomy along a defined spectrum (from command-driven to consensus to locally autonomous/master) for each goal it pursues. This article presents one motivation for DAA through experiments showing that the ‘best’ decision-making framework for a group of agents depends not only on the problem domain and pre-defined characteristics of the system, but also on run-time factors that can change during system operation. This result holds regardless of which performance metric is used to define ‘best’. Thus, it is possible for agents to benefit by dynamically adapting their decision-making frameworks to their situation during system operation.