scholarly journals Efficiently Combining Human Demonstrations and Interventions for Safe Training of Autonomous Systems in Real-Time

2019 ◽  
Vol 33 ◽  
pp. 2462-2470 ◽  
Author(s):  
Vinicius G. Goecks ◽  
Gregory M. Gremillion ◽  
Vernon J. Lawhern ◽  
John Valasek ◽  
Nicholas R. Waytowich
Keyword(s):  
Autonomous Systems ◽  
Human Interaction ◽  
Task Completion ◽  
Autonomous Agent ◽  
High Fidelity ◽  
Current Effort ◽  
Multiple Modalities ◽  
Simulated Environment ◽  
Learning From Demonstrations ◽  
Overall Performance

This paper investigates how to utilize different forms of human interaction to safely train autonomous systems in realtime by learning from both human demonstrations and interventions. We implement two components of the Cycle-of Learning for Autonomous Systems, which is our framework for combining multiple modalities of human interaction. The current effort employs human demonstrations to teach a desired behavior via imitation learning, then leverages intervention data to correct for undesired behaviors produced by the imitation learner to teach novel tasks to an autonomous agent safely, after only minutes of training. We demonstrate this method in an autonomous perching task using a quadrotor with continuous roll, pitch, yaw, and throttle commands and imagery captured from a downward-facing camera in a high-fidelity simulated environment. Our method improves task completion performance for the same amount of human interaction when compared to learning from demonstrations alone, while also requiring on average 32% less data to achieve that performance. This provides evidence that combining multiple modes of human interaction can increase both the training speed and overall performance of policies for autonomous systems.

scholarly journals A Human-Inspired Control Strategy for Improving Seamless Robot-To-Human Handovers

2021 ◽  
Vol 11 (10) ◽  
pp. 4437
Author(s):  
Paramin Neranon ◽  
Tanapong Sutiphotinun
Keyword(s):  
Control System ◽  
Control Strategy ◽  
Impedance Control ◽  
Human Interaction ◽  
Robotic Control ◽  
Natural Manner ◽  
Object Handover ◽  
Overall Performance ◽  
Survey Responses ◽  
Test Outcomes

One of the challenging aspects of robotics research is to successfully establish a human-like behavioural control strategy for human–robot handover, since a robotic controller is further complicated by the dynamic nature of the human response. This paper consequently highlights the development of an appropriate set of behaviour-based control for robot-to-human object handover by first understanding an equivalent human–human handover. The optimized hybrid position and impedance control was implemented to ensure good stability, adaptability and comfort of the robot in the object handover tasks. Moreover, a questionnaire technique was employed to gather information from the participants concerning their evaluations of the developed control system. The results demonstrate that the quantitative measurement of performance of the human-inspired control strategy can be considered acceptable for seamless human–robot handovers. This also provided significant satisfaction with the overall control performance in the robotic control system, in which the robot can dexterously pass the object to the receiver in a timely and natural manner without the risk of harm or injury by the robot. Furthermore, the survey responses were in agreement with the parallel test outcomes, demonstrating significant satisfaction with the overall performance of the robot–human interaction, as measured by an average rating of 4.20 on a five-point scale.

scholarly journals Long-Term Adaptivity in Distributed Intelligent Systems: Study of ViaBots in a Simulated Environment

Robotics ◽  
2019 ◽  
Vol 8 (2) ◽  
pp. 25
Author(s):  
Arturs Ardavs ◽  
Mara Pudane ◽  
Egons Lavendelis ◽  
Agris Nikitenko
Keyword(s):  
Organization Theory ◽  
Intelligent Systems ◽  
Autonomous Robots ◽  
Autonomous Agent ◽  
Systems Model ◽  
Simulated Environment ◽  
Multi Agent ◽  
Distributed Intelligent Systems ◽  
Viable System

This paper proposes a long-term adaptive distributed intelligent systems model which combines an organization theory and multi-agent paradigm—ViaBots. Currently, the need for adaptivity in autonomous intelligent systems becomes crucial due to the increase in the complexity and diversity of the tasks that autonomous robots are employed for. To deal with the design complexity of such systems within the ViaBots model, each part of the modeled system is designed as an autonomous agent and the entire model, as a multi-agent system. Based on the viable system model, which is widely used to ensure viability, (i.e., long-term autonomy of organizations), the ViaBots model defines the necessary roles a system must fulfill to be capable to adapt both to changes in its environment (like changes in the task) and changes within the system itself (like availability of a particular robot). Along with static role assignments, ViaBots propose a mechanism for role transition from one agent to another as one of the key elements of long term adaptivity. The model has been validated in a simulated environment using an example of a conveyor system. The simulated model enabled the multi-robot system to adapt to the quantity and characteristics of the available robots, as well as to the changes in the parts to be processed by the system.

scholarly journals A Fuzzy-Logic Autonomous Agent Applied as a Supervisory Controller in a Simulated Environment

2004 ◽  
Vol 12 (1) ◽  
pp. 107-122 ◽  
Author(s):  
G. Chrysanthakopoulos ◽  
W.L.J. Fox ◽  
R.T. Miyamoto ◽  
R.J. Marks ◽  
M.A. El-Sharkawi ◽  
...  
Keyword(s):  
Fuzzy Logic ◽  
Autonomous Agent ◽  
Supervisory Controller ◽  
Simulated Environment

scholarly journals Comparative study of user experience on mobile pedestrian navigation between digital map interface and location-based augmented reality

2020 ◽  
Vol 10 (2) ◽  
pp. 2037
Author(s):  
Komang Candra Brata ◽  
Deron Liang
Keyword(s):  
Augmented Reality ◽  
User Experience ◽  
Technology Development ◽  
Task Completion ◽  
Navigation Systems ◽  
Digital Map ◽  
Pedestrian Navigation ◽  
Points Of Interest ◽  
Overall Performance ◽  
Hedonic Quality

Fast-paced mobile technology development has permitted augmented reality experiences to be delivered on mobile pedestrian navigation context. The fact that the more prevalent of this technology commonly will substituting the digital map visualization to present the geo-location information is still debatable. This paper comprises a report on a field study comparing about user experience when interacting with different modes of mobile electronic assistance in the context of pedestrian navigation interfaces which utilize location-based augmented reality (AR) and two-dimensional digital map to visualize the points of interest (POIs) location in the vicinity of the user. The study was conducted with two subsequent experiments in the Zhongli District, Taoyuan City, Taiwan. The study involved 10 participants aged between 22 and 28 years with different experiences in using smartphones and navigation systems. Navigation performance was measured based on a usability approach on pragmatic quality and hedonic quality like effectiveness (success rate of task completion), efficiency (task completion time) and satisfaction in real outdoor conditions. The evaluation findings have been cross-checked with the user’s personal comments. We aim at eliciting knowledge about user requirements related to mobile pedestrian interfaces and evaluating user experience from pragmatic and hedonic viewpoints. Results show that in the context of pedestrian navigation, digital map interfaces lead to significantly better navigation performance in pragmatic attributes in comparison to AR interfaces. Nevertheless, the study also reveals that location-based AR is more valued by participants in hedonic qualities and overall performance.

Machine Learning for Evaluating the Social Impact of Engineered Products: A Framework

2019 ◽  
Author(s):  
Bryan J. Stringham ◽  
Daniel O. Smith ◽  
Christopher A. Mattson ◽  
Eric C. Dahlin
Keyword(s):  
Machine Learning ◽  
Social Impact ◽  
Low Frequency ◽  
Human Interaction ◽  
High Fidelity ◽  
Frequency Data ◽  
Hand Pump ◽  
The Social ◽  
Impact Indicators ◽  
Water Pumps

Abstract Evaluating the social impact indicators of engineered products is crucial to better understanding how products affect individuals’ lives and discover how to design for positive social impact. Most existing methods for evaluating social impact indicators require direct human interaction with users of a product, such as one-on-one interviews. These interactions produce high-fidelity data that are rich in information but provide only a single snapshot in time of the product’s impacts and are less frequently collected due to the significant human resources and cost associated with obtaining them. A framework is proposed that describes how low-fidelity data passively obtained using remote sensors, satellites, and digital technology can be collected and correlated with high-fidelity, low-frequency data using machine learning. Using this framework provides an inexpensive way to continuously monitor the social impact indicators of products by augmenting high-fidelity, low-frequency data with low-fidelity, continuously-collected data using machine learning. We illustrate an application of this framework by demonstrating how it can be used to examine the gender-related social impact indicators of water pumps in Uganda. The provided example uses a deep learning model to correlate pump handle movement (measured via an integrated motion unit) with user type (man, woman, or child) of 1,200 hand pump users.

The Humanity of Humanless Systems

2019 ◽  
Vol 28 (3) ◽  
pp. 4-6 ◽  
Author(s):  
Peter A. Hancock
Keyword(s):  
Quality Of Life ◽  
Cycle Time ◽  
Information Exchange ◽  
Autonomous Systems ◽  
Human Interaction ◽  
Machine Cycle ◽  
Collaborative Interaction ◽  
Human Quality ◽  
Information Interchange

This work considers the future of human interaction with progressively more autonomous systems. I argue that the temporal dissonance between the human’s ‘cycle time’ and machine ‘cycle time,’ will become an overwhelming barrier to collaborative interaction. We may slow machines, we may buffer information exchange, we may default to meta-levels of strategic interchange but in the end all transparency of information interchange will dissolve under the driving influence of time. HF/E is thus already fighting rear-guard action. The question remains as to the sustenance of human quality of life in this evolving milieu.

Reaction Times When Switching From Autonomous to Manual Driving Control

2016 ◽  
Vol 60 (1) ◽  
pp. 1854-1858 ◽  
Author(s):  
Kelly Funkhouser ◽  
Frank Drews
Keyword(s):  
Autonomous Vehicles ◽  
Human Performance ◽  
Autonomous Systems ◽  
Reaction Times ◽  
Cognitive Engagement ◽  
Human Interaction ◽  
Transitional State ◽  
The Moment ◽  
Future Technologies ◽  
Current Reality

As autonomous vehicles become more prevalent in our everyday lives, we must succumb to the realities of technological deficiencies. Although a future of fully autonomous vehicles would be the pinnacle of safety and efficiency, the current reality leaves us in a transitional state requiring human interaction with autonomous systems. Therefore it is imperative to understand human-system interaction with the autonomous features in current and future technologies. To gain an improved understanding, we designed an investigational study to gain a better understanding of human performance parameters at the moment they relieve and regain control of autonomous systems. The current findings show that reaction time increases as time disengaged from the task of driving increases, regardless of cognitive engagement.

scholarly journals A novel audio based human interaction proof for visually challenged users

2018 ◽  
Vol 7 (2.7) ◽  
pp. 289
Author(s):  
Rishi Korrapolu ◽  
Manoj Sai. N ◽  
Kameshwara Rao.M
Keyword(s):  
Instrumental Music ◽  
Human Interaction ◽  
Huge Number ◽  
Overall Performance ◽  
Visually Challenged ◽  
The Web

CAPTCHAs are strategies to recognize human clients and PC programs naturally. CAPTCHAs shield different sorts of online administrations from beast compel assaults and foreswearing of administration via programmed PC programs. Most CAPTCHAs comprise of mages with misshaped content. Shockingly, visual CAPTCHAs constrain access to the a huge number of outwardly hindered individuals utilizing the Web. Sound CAPTCHAs were made to fathom this openness issue. However the presently accessible sound CAPTCHAs have been broken with differing achievement, utilizing the shortcoming in the techniques utilized. Our system, presents the user with an interface that plays a song using instrumental music (nonvocal) randomly selected from some language of users choice. The user is then asked to kind the music composer and then the device estimates whether it is a human or no longer by means of analyzing the response. A person look at turned into conducted to research the overall performance of our proposed mechanism.

An Investigation of Team Inputs, Processes, And Emergent States on Performance in a Spaceship Bridge Simulation

2021 ◽  
Vol 65 (1) ◽  
pp. 1475-1479
Author(s):  
Richard J. Simonson ◽  
Joseph R. Keebler ◽  
Ryan J. Wallace ◽  
Andrew C. Griggs
Keyword(s):  
Team Performance ◽  
Systems Engineering ◽  
Team Effectiveness ◽  
Team Processes ◽  
Simulated Environment ◽  
Emergent States ◽  
Overall Performance ◽  
Input Variables ◽  
Research Initiatives ◽  
Individual Input

This investigation serves to provide evidence on the effects of various input variables on intact teams through repeated team performance sessions in a team spaceship bridge simulation (i.e. Artemis). The Input Mediator/Moderator Output Input (IMOI) model provides a systems engineering an approach to understand various team and individual input variables contribution to the development of team processes and emergent states, ultimately leading to a team’s ability to perform together. While various prior research initiatives have served to contribute to the pool of evidence of which input variables are most highly predictive of a team’s overall performance, the need for further and recursive input to output investigations is needed. Our results indicate perceived team effectiveness and cohesion are significant predictors in team performance and that skill and knowledge of a simulated environment may overshadow team-specific effectiveness indicators as the team gains experience.

Sign in / Sign up

Export Citation Format

Share Document