{"605828":{"#nid":"605828","#data":{"type":"event","title":"Ph.D. Proposal Oral Exam - Matthew O\u2019Brien","body":[{"value":"\u003Cp\u003E\u003Cstrong\u003ETitle:\u003C\/strong\u003E Entraining a Robot to its Environment with an Artificial Circadian System\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u003Cstrong\u003ECommittee:\u003C\/strong\u003E\u003C\/p\u003E\r\n\r\n\u003Cp\u003EDr. Ronald Arkin (advisor)\u003C\/p\u003E\r\n\r\n\u003Cp\u003EDr. Magnus Egerstedt\u003C\/p\u003E\r\n\r\n\u003Cp\u003EDr. Ayanna Howard\u003C\/p\u003E\r\n\r\n\u003Cp\u003EDr. C\u0026eacute;dric Pradalier\u003C\/p\u003E\r\n\r\n\u003Cp\u003EDr. Zsolt Kira\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u003Cstrong\u003EAbstract:\u003C\/strong\u003E\u003C\/p\u003E\r\n\r\n\u003Cp\u003EAs robots become persistent agents in the world outside of factory floors, they must deal with continuously changing environments. This challenge has grown research in long-term autonomy, with particular focus on localization and mapping in dynamic environments. Less attention has been paid to learning these dynamics to adapt or entrain an agent\u0026rsquo;s behavior to them.\u003C\/p\u003E\r\n\r\n\u003Cp\u003EInspired by circadian systems in nature, this proposal seeks to answer how a robotic agent can both learn the regular cycles and patterns that exist in many environments (e.g. sunlight, traffic, etc), and how it can exploit that knowledge. The core approach is to model relevant environmental states as time series. These models are treated as special percepts in a behavior-based architecture; providing forecasts of the future state, rather than measurements of the current state. Action-selection is accomplished using behavior-specific activation functions, which like the behavior\u0026rsquo;s themselves, incorporate both the measured and predicted state.\u003C\/p\u003E\r\n\r\n\u003Cp\u003EThe proposed work centers around developing and validating this approach on a physical robot interacting with a dynamic environment. In addition, two theoretical extensions are considered: making the approach robust to random events and inevitable model failure, and investigating how reinforcement learning may be applied as an alternative method for action-selection.\u003C\/p\u003E\r\n","summary":null,"format":"limited_html"}],"field_subtitle":"","field_summary":"","field_summary_sentence":[{"value":"Entraining a Robot to its Environment with an Artificial Circadian System"}],"uid":"28475","created_gmt":"2018-05-03 21:14:48","changed_gmt":"2018-05-03 21:14:48","author":"Daniela Staiculescu","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2018-05-16T18:00:00-04:00","event_time_end":"2018-05-16T20:00:00-04:00","event_time_end_last":"2018-05-16T20:00:00-04:00","gmt_time_start":"2018-05-16 22:00:00","gmt_time_end":"2018-05-17 00:00:00","gmt_time_end_last":"2018-05-17 00:00:00","rrule":null,"timezone":"America\/New_York"},"extras":[],"groups":[{"id":"434371","name":"ECE Ph.D. Proposal Oral Exams"}],"categories":[],"keywords":[{"id":"102851","name":"Phd proposal"},{"id":"1808","name":"graduate students"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1788","name":"Other\/Miscellaneous"}],"invited_audience":[{"id":"78771","name":"Public"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}