Modeling human expertise for providing adaptive levels of robot shared autonomy Public Deposited

http://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/2b88qh98s

Descriptions

Attribute NameValues
Creator
Abstract or Summary
  • In shared autonomy, a robot and human user both have some level of control in order to achieve a shared goal. Choosing the balance of control given to the user and the robot can be a challenging problem since different users have different preferences and vary in skill levels when operating a robot. We propose using a novel formulation of Partially Observable Markov Decision Processes (POMDPs) to represent a model of the user's expertise in controlling the robot. The POMDP uses observations from the user's actions and from the environment to update the belief of the user's skill and chooses a level of control between the robot and the user. The level of control given between the user and the robot is encapsulated in macro-action controllers. The macro-action controllers encompass varying levels of robot autonomy and reduce the space of the POMDP, removing the need to plan over separate actions. As part of this research, we ran two users study, developed a method to automatically generate macro-action controller values, and applied our user expertise model to provide shared autonomy on a semi-autonomous underwater vehicle. In our first user study, we tested our user expertise model in a robot driving simulation. Users drove a simulated robot through an obstacle-filled map while the POMDP model chose appropriate macro-action controllers based on the belief state of the user's skill level. The results of the user study showed that our model can encapsulate user skill levels. The results also showed that using the controller with greater robot autonomy helped users of low skill avoid obstacles more than it helped users of high skill. We designed a controller value synthesis method to generate the variables that control the levels of autonomy in the macro-action controllers. We found differences in how the users drive the robot using a decision tree generated from the data recorded in the first user study, and we used these differences to program simulated user ``bots'' that mimic users of different skill levels. The ``bots'' were used to test a range of variables for the controllers, and the controller variables were found from minimizing obstacles hit, time to complete maps, and total distance driven from the simulated data.For our second user study, we looked at users' satisfaction without robot autonomy, with the highest amount of autonomy, and with the autonomy chosen by our expertise model. We found users we classified as beginners ranked the autonomy more favorably than those ranked as experts. We implemented our expertise model on a Seabotix vLBV300 underwater vehicle and ran a trial off the coast of Newport, Oregon. During our trials, we recorded a user driving the vehicle to predetermined waypoints. When beginner actions were performed, the user expertise model provided an increased level of autonomy which either increased throttle when far from waypoints or decreased throttle when close to waypoints. This demonstrated an implementation of our algorithm on existing robot hardware in the field.
License
Resource Type
Date Available
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Non-Academic Affiliation
Keyword
Rights Statement
Publisher
Peer Reviewed
Language
Replaces
Additional Information
  • description.provenance : Approved for entry into archive by Margaret Mellinger(margaret.mellinger@oregonstate.edu) on 2017-10-16T23:56:24Z (GMT) No. of bitstreams: 2license_rdf: 1536 bytes, checksum: df76b173e7954a20718100d078b240a8 (MD5)MillikenLaurenM2018.pdf: 3220276 bytes, checksum: 52793d25f19c2dae768b8fbae850fe05 (MD5)
  • description.provenance : Approved for entry into archive by Julie Kurtz(julie.kurtz@oregonstate.edu) on 2017-09-25T18:58:07Z (GMT) No. of bitstreams: 2license_rdf: 1536 bytes, checksum: df76b173e7954a20718100d078b240a8 (MD5)MillikenLaurenM2018.pdf: 3220276 bytes, checksum: 52793d25f19c2dae768b8fbae850fe05 (MD5)
  • description.provenance : Submitted by Lauren Milliken (millikel@oregonstate.edu) on 2017-09-19T22:08:45ZNo. of bitstreams: 2license_rdf: 1536 bytes, checksum: df76b173e7954a20718100d078b240a8 (MD5)MillikenLaurenM2018.pdf: 3220276 bytes, checksum: 52793d25f19c2dae768b8fbae850fe05 (MD5)
  • description.provenance : Made available in DSpace on 2017-10-16T23:56:24Z (GMT). No. of bitstreams: 2license_rdf: 1536 bytes, checksum: df76b173e7954a20718100d078b240a8 (MD5)MillikenLaurenM2018.pdf: 3220276 bytes, checksum: 52793d25f19c2dae768b8fbae850fe05 (MD5) Previous issue date: 2017-08-18

Relationships

Parents:

This work has no parents.

Last modified

Downloadable Content

Download PDF

Items