{"657942":{"#nid":"657942","#data":{"type":"news","title":"New Framework for Cooperative Bots Mimics High-Functioning Human Teams, Decreases Risks from Unreliable Bots","body":[{"value":"\u003Cp\u003EA Georgia Institute of Technology\u0026nbsp;research group in the School of Interactive Computing has developed a robotics system that exceeds existing standards for collaborative bots that work independently to achieve a shared goal.\u003C\/p\u003E\r\n\r\n\u003Cp\u003EThe system intelligently increases the information shared among the bots and \u003Ca href=\u0022https:\/\/www.youtube.com\/watch?v=rK_itCF9hPc\u0022\u003Eallows for improved cooperation\u003C\/a\u003E. The aim is to model high-functioning human teams. It also creates resiliency against bad or unreliable team bots that may hinder the overall programmed goal.\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u0026ldquo;Intuitively, the idea behind our new framework \u0026mdash; InfoPG \u0026mdash;\u0026nbsp;is that a robot agent goes back-and-forth on what it thinks it \u003Cem\u003Eshould\u003C\/em\u003E do with their teammates, and then the teammates will update on what they think is \u003Cem\u003Ebest\u003C\/em\u003E to do,\u0026rdquo; said \u003Cstrong\u003EEsmaeil Seraj\u003C\/strong\u003E, Ph.D. student in the \u003Ca href=\u0022https:\/\/core-robotics.gatech.edu\/\u0022\u003ECORE Robotics Lab\u003C\/a\u003E and researcher on the project. \u0026ldquo;They do this until the decision is deeply rationalized and reasoned about.\u0026rdquo;\u003C\/p\u003E\r\n\r\n\u003Cp\u003EThe work focuses on artificial agents on a decentralized team \u0026mdash; in simulations or the real world \u0026mdash;\u0026nbsp;working in concert toward a specific task. Applications could include surgery, search and rescue, and disaster response, among others.\u003C\/p\u003E\r\n\r\n\u003Cp\u003EInfoPG facilitates communication between the artificial agents on an iterative basis and allows for actions and decisions that mimic human teams working at optimal levels.\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u0026ldquo;This research is in fact inspired by how high-performing human teams act,\u0026rdquo; said Seraj.\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u0026ldquo;Humans normally use k-level thinking \u0026mdash; such as, \u0026lsquo;what I think you will do, what I think you think I will do, and so on\u0026rsquo; \u0026mdash; to rationalize their actions in a team,\u0026rdquo; he said. \u0026ldquo;The basic thought is that the more you know about your teammate\u0026#39;s strategy, the easier it is for you to take the best action possible.\u0026rdquo;\u003C\/p\u003E\r\n\r\n\u003Cp\u003EUsing this approach, the researchers designed InfoPG to make one bot\u0026rsquo;s decisions conditional on its teammates. They ran simulations using simple games like Pong, and complex games like StarCraft II.\u003C\/p\u003E\r\n\r\n\u003Cp\u003EIn the latter \u0026mdash; where the goal is for one team of agents to defeat another \u0026mdash; the InfoPG architecture showed very advanced strategies. Seraj said agents in one case learned to form a triangle formation, sacrificing the front agent while the two other agents eliminated the enemy. Without InfoPG in play, an agent abandoned its team to save itself.\u003C\/p\u003E\r\n\r\n\u003Cp\u003EThe new method also limits the disruption a bad bot on the team might cause.\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u0026ldquo;Coordinating actions with such a fraudulent agent in a collaborative multi-agent setting can be detrimental,\u0026rdquo; said \u003Cstrong\u003EMatthew Gombolay\u003C\/strong\u003E, assistant professor in the School of Interactive Computing and director of the CORE Robotics Lab. \u0026ldquo;We need to ensure the integrity of robot teams in real-world applications where bots might be tasked to save lives or help people and organizations extend their capabilities.\u0026rdquo;\u003C\/p\u003E\r\n\r\n\u003Cp\u003EResults of the work show InfoPG\u0026rsquo;s performance exceeds various baselines in learning cooperative policies for multi-agent reinforcement learning. The researchers plan to move the system from simulation into real robots, such as controlling a swarm of drones to help surveil and fight wildfires.\u003C\/p\u003E\r\n\r\n\u003Cp\u003EThe research is published in the 2022 Proceedings of the International Conference on Learning Representations. The paper, \u003Cem\u003EIterated Reasoning with Mutual Information in Cooperative and Byzantine Decentralized Teaming\u003C\/em\u003E is co-authored by computer science major \u003Cstrong\u003ESachin G. Konan\u003C\/strong\u003E, Seraj, and Gombolay.\u003C\/p\u003E\r\n\r\n\u003Cp\u003EThis work was sponsored by the Office of Naval Research under grant N00014-19-1-2076 and the Naval Research Lab (NRL) under the grant N00173-20-1-G009. The researchers\u0026rsquo; views and statements are based on their findings and do not necessarily reflect those of the funding agencies.\u003C\/p\u003E\r\n","summary":null,"format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003EA Georgia Tech research group in the School of Interactive Computing has developed a robotics system that exceeds existing standards for collaborative bots that work independently to achieve a shared goal.\u003C\/p\u003E\r\n","format":"limited_html"}],"field_summary_sentence":[{"value":"A Georgia Tech research group in the School of Interactive Computing has developed a robotics system that exceeds existing standards for collaborative bots that work independently to achieve a shared goal. "}],"uid":"27592","created_gmt":"2022-05-04 13:53:03","changed_gmt":"2022-05-05 18:12:26","author":"Joshua Preston","boilerplate_text":"","field_publication":"","field_article_url":"","dateline":{"date":"2022-05-04T00:00:00-04:00","iso_date":"2022-05-04T00:00:00-04:00","tz":"America\/New_York"},"extras":[],"hg_media":{"657945":{"id":"657945","type":"image","title":"Improving collaboration in decentralized teams of bots","body":null,"created":"1651672727","gmt_created":"2022-05-04 13:58:47","changed":"1651672727","gmt_changed":"2022-05-04 13:58:47","alt":"","file":{"fid":"249393","name":"promo_graphic_ICLR22_collab robots.jpg","image_path":"\/sites\/default\/files\/images\/promo_graphic_ICLR22_collab%20robots.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/images\/promo_graphic_ICLR22_collab%20robots.jpg","mime":"image\/jpeg","size":589661,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/images\/promo_graphic_ICLR22_collab%20robots.jpg?itok=IZixdgDo"}},"657946":{"id":"657946","type":"image","title":"Multiwalker bots coordinating to carry object","body":null,"created":"1651672836","gmt_created":"2022-05-04 14:00:36","changed":"1651672836","gmt_changed":"2022-05-04 14:00:36","alt":"","file":{"fid":"249394","name":"robot cooperation in decentralized team.png","image_path":"\/sites\/default\/files\/images\/robot%20cooperation%20in%20decentralized%20team.png","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/images\/robot%20cooperation%20in%20decentralized%20team.png","mime":"image\/png","size":2326733,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/images\/robot%20cooperation%20in%20decentralized%20team.png?itok=VL1FfTYk"}}},"media_ids":["657945","657946"],"groups":[{"id":"576481","name":"ML@GT"},{"id":"50876","name":"School of Interactive Computing"},{"id":"1188","name":"Research Horizons"}],"categories":[],"keywords":[{"id":"187915","name":"go-researchnews"}],"core_research_areas":[{"id":"39521","name":"Robotics"}],"news_room_topics":[],"event_categories":[],"invited_audience":[],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[{"value":"\u003Cp\u003E\u003Ca href=\u0022mailto:jpreston7@gatech.edu\u0022\u003EJoshua Preston\u003C\/a\u003E\u003Cbr \/\u003E\r\nResearch Communications Manager\u003Cbr \/\u003E\r\nCollege of Computing\u003C\/p\u003E\r\n","format":"limited_html"}],"email":["jpreston7@gatech.edu"],"slides":[],"orientation":[],"userdata":""}}}