Designing Information Gathering Robots for Human-Populated Environments

Abstract - Advances in mobile robotics have enabled robotsthat can autonomously operate in human-populated environments. Although primary tasks for such robots might be fetching, delivery, or escorting, they present an untapped potentialas information gathering agents that can answer questions forthe community of co-inhabitants. In this paper, we seek tobetter understand requirements for such information gatheringrobots (InfoBots) from the perspective of the user requestingthe information. We present findings from two studies: (i) a usersurvey conducted in two office buildings and (ii) a 4-day longdeployment in one of the buildings, during which inhabitantsof the building could ask questions to an InfoBot through aweb-based interface. These studies allow us to characterize thetypes of information that InfoBots can provide for their users.

For details, see https://www.researchgate.net/publication/308819532_Designing_information_gathering_robots_for_human-populated_environments

Transcript

1. Designing Information Gathering Robots for Human-Populated Environments IROS 2015 Maya Cakmak Rajesh Rao UNIVERSITY OF WASHINGTON Mike Chung Andrzej Pronobis Dieter Fox
2. Hello there
3. Hello there Aethon TUG Savioke Relay CMU Cobots
4. Is my advisor in her office? Can you find me an empty study lounge? Hello there Let me know when my advisor becomes available What was your day like?
5. INFOBOTS • Information gathering agents in human populated environments We seek to better understand the value of the services that InfoBots can provide
6. Is my advisor in her office? Can you find me an empty study lounge? Let me know when my advisor becomes available What was your day like? Checking
 Searching
 Monitoring
 Summarizing TASK TYPES FOR INFOBOTS
7. Is my advisor in her office? Can you find me an empty study lounge? Let me know when my advisor becomes available What was your day like? Checking
 Searching
 Monitoring
 Summarizing TASK TYPES FOR INFOBOTS
8. STUDY DESIGN Study 1: A User Survey
 • to determine types of useful information 
 • to identify constraints and requirements Study 2: Deployment
 • to study practical usage of the service
9. STUDY DESIGN Study 1: A User Survey
 • to determine types of useful information 
 • to identify constraints and requirements Study 2: Deployment
 • to study practical usage of the service
10. Q1. Is John in his office? Q2. How many people are in the lounge Q3. Are there any empty tables in the study 
 room? Q4. Are there any bagels at the coffeeshop? Q5. Is there free food in the kitchen? Q6. Is the conference room occupied?
11. FINDINGS: USEFULNESS Very useful Useful Good to know Not so useful Useless 15% 30% 45% 60% 15% 30% 45% 60% CS LAW Average % of responses over 6 questions The ability to ask this type of question would be
12. FINDINGS: USEFULNESS Very useful Useful Good to know Not so useful Useless 15% 30% 45% 60% 15% 30% 45% 60% CS LAW Average % of responses over 6 questions 84% 79% People think InfoBots can be useful The ability to ask this type of question would be
13. FINDINGS: USAGE FREQUENCY Multiple times a day Every day Once/twice a week Once/twice a month Never 15% 30% 45% 60% 15% 30% 45% 60% CS LAW Average % of responses over 6 questions People will not ask questions frequently I would ask this type of question
14. FINDINGS: USEFULNESS Very useful Useful Good to know Not so useful Useless 15% 30% 45% 60% 15% 30% 45% 60% CS LAW Average % of responses over 6 questions 84% 79% People think InfoBots can be useful The ability to ask this type of question would be
15. FINDINGS: USAGE FREQUENCY Multiple times a day Every day Once/twice a week Once/twice a month Never 15% 30% 45% 60% 15% 30% 45% 60% CS LAW Average % of responses over 6 questions People will not ask questions frequently 25% 21% I would ask this type of question
16. FINDINGS: TIME CONSTRAINT Immediatedly Faster than a human can Same speed as a human Half the speed of a human No rush whenever 15% 30% 45% 60% 15% 30% 45% 60% CS LAW Average % of responses over 6 questions 79% 79% High expectation I would require a response
17. STUDY DESIGN Study 1: A User Survey
 • to determine types of useful information 
 • to identify constraints and requirements Study 2: Deployment
 • to study practical usage of the service
23. Yes.
24. FINDINGS: QUESTION TYPES Is there anyone in {location}? checking Is {person} in his/her office? checking Is there any food in the downstairs kitchen? checking Is there anything in my mailbox? checking Does {name}’s office have a sofa? checking Is the reception still open? checking Who let the dogs out? :) non-checking Has {person} arrived yet today in the {} building? non-checking Which meeting room has the best visibility of the {landmark} today? non-checking 0 20 40 60 80 80% 20% Checking questions Non-checking questions number of questions
25. FINDINGS: CHECKING SUBTYPES Is there anyone in {location}? presence Is {person} in his/her office? presence Is there any food in the downstairs kitchen? presence Is there anything in my mailbox? presence Does {name}’s office have a sofa? presence Is the reception still open? state 0 20 40 60 80 76% 24% Presence questions State questions number of questions =0.89 e.g. Is {object} in {location}? e.g. Is {location} {state}? presence: state:
26. FINDINGS: TARGET OBJECTS Is there anyone in {location}? person Is {person} in his/her office? person Is there any food in the downstairs kitchen? food Is there anything in my mailbox? mail Does {name}’s office have a sofa? other 0 20 40 60 80 60% 21% Person Food Mail Other 6% 13% number of questions Long tailed distribution =0.93
27. POST-DEPLOYMENT SURVEY Please rate how satisfied you were with InfoBot's *speed* in answering your question. Completely satisfied Very satisfied Moderately satisfied Slightly satisfied Not at all satisfied 15% 30% 45% 60% 5% Despite initial high expectation,
 only 5% was “Not at all satisfied”.
28. PROGRESS ON AUTONOMOUS INFOBOTS Q. Is there breakout area occupied? whiteboard presence( cse 400 , person ) Submitted to ICRA2016 Input: Output: language parsing viewpoint estimation
29. CONTRIBUTIONS 1. Categorization of InfoBots’ task types 2. Findings on people expected InfoBot usage 3. Findings on actual InfoBot usage Thank you!