BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:Australia/Adelaide BEGIN:DAYLIGHT DTSTART:20221002T030000 TZOFFSETFROM:+0930 TZOFFSETTO:+1030 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=10 TZNAME:ACDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20220403T020000 TZOFFSETFROM:+1030 TZOFFSETTO:+0930 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=4 TZNAME:ACST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20220624T110823Z UID:10865795-96EF-464C-9BF6-C3808A16CCBF DTSTART;TZID=Australia/Adelaide:20220621T173000 DTEND;TZID=Australia/Adelaide:20220621T183000 DESCRIPTION:(Hybrid Seminar)\n\nPlease indicate if you are attending phyisc ally or virtually when you are registering.\n\nAbstract\n\nCoordinated mul ti-robot systems are an effective way to harvest data from sensor networks and to implement active perception strategies. However\, achieving effici ent coordination in a way which guarantees a target QoS while adapting dyn amically to changes (in the environment\, due to sensors’ mobility\, and /or in the value of harvested data) is to date a key open issue. In this w ork\, we propose a novel decentralized Monte Carlo Tree Search algorithm ( MCTS) which allows agents to optimize their own actions while achieving so me form of coordination\, in a changing environment. Its key underlying id ea is to balance in an adaptive manner the exploration-exploitation trade- off to deal effectively with abrupt changes caused by the environment and random changes caused by other agents' actions. Critically\, outdated and irrelevant samples - an inherent and prevalent feature in all multi-agent MCTS-based algorithms - are filtered out by means of a sliding window mech anism. We show both theoretically and through simulations that our algorit hm provides a log-factor (in terms of time steps) smaller regret than stat e-of-the-art decentralized multi-agent planning methods. We instantiate ou r approach on the problem of underwater data collection\, showing on a set of different models for changes that our approach greatly outperforms the best available algorithms for that setting\, both in terms of convergence speed and of global utility.\n\nNote\n\nThe seminar is hybrid. Physical r oom location will be provided after registration prior to seminar day.\n\n Light drinks and snacks will be provided after the seminar for networking opportunities.\n\nThe seminar will also be available via Zoom.\n\nIf you a re unable to attend physically and wish to participate please register and you will receive an email with the details in the lead up to the event.\n \nFurther information\n\nPlease contact:\n\nadriel@ieee.org\n\nkyle.millar @ieee.org\n\nSpeaker(s): Nhat Nguyen\, \n\nRoom: EMG06 or Zoom\, Bldg: Eng ineering and Maths Building\, The University of Adelaide\, North Terrace C ampus\, https://adelaide.zoom.us/j/84464256298?pwd=RDRVZ2RNM1Q5aTZOaGVuMTF HeVZadz09\, ADELAIDE\, South Australia\, Australia\, 5005\, Virtual: https ://events.vtools.ieee.org/m/316656 LOCATION:Room: EMG06 or Zoom\, Bldg: Engineering and Maths Building\, The U niversity of Adelaide\, North Terrace Campus\, https://adelaide.zoom.us/j/ 84464256298?pwd=RDRVZ2RNM1Q5aTZOaGVuMTFHeVZadz09\, ADELAIDE\, South Austra lia\, Australia\, 5005\, Virtual: https://events.vtools.ieee.org/m/316656 ORGANIZER:kyle.millar@ieee.org SEQUENCE:8 SUMMARY:Multi-Agent Data Collection in Non-Stationary Environments URL;VALUE=URI:https://events.vtools.ieee.org/m/316656 X-ALT-DESC:Description: <br /><h2>(Hybrid Seminar)</h2>\n<p>Please indicate if you are attending phyiscally or virtually when you are registering.</p >\n<h2>Abstract</h2>\n<p>Coordinated multi-robot systems are an effective way to harvest data from sensor networks and to implement active perceptio n strategies. However\, achieving efficient coordination in a way which gu arantees a target QoS while adapting dynamically to changes (in the enviro nment\, due to sensors&rsquo\; mobility\, and/or in the value of harvested data) is to date a key open issue. In this work\, we propose a novel dece ntralized Monte Carlo Tree Search algorithm (MCTS) which allows agents to optimize their own actions while achieving some form of coordination\, in a changing environment. Its key underlying idea is to balance in an adapti ve manner the exploration-exploitation trade-off to deal effectively with abrupt changes caused by the environment and random changes caused by othe r agents' actions. Critically\, outdated and irrelevant samples - an inher ent and prevalent feature in all multi-agent MCTS-based algorithms - are f iltered out by means of a sliding window mechanism. We show both theoretic ally and through simulations that our algorithm provides a log-factor (in terms of time steps) smaller regret than state-of-the-art decentralized mu lti-agent planning methods. We instantiate our approach on the problem of underwater data collection\, showing on a set of different models for chan ges that our approach greatly outperforms the best available algorithms fo r that setting\, both in terms of convergence speed and of global utility. </p>\n<h3>Note</h3>\n<p>The seminar is hybrid. Physical room location will be provided after registration prior to seminar day.&nbsp\;</p>\n<p>Light drinks and snacks will be provided after the seminar for networking oppor tunities.</p>\n<p>The seminar will also be available via Zoom.</p>\n<p>If you are unable to attend physically and wish to participate please registe r and you will receive an email with the details in the lead up to the eve nt.</p>\n<h2>Further information</h2>\n<p>Please contact:</p>\n<p>adriel@i eee.org</p>\n<p>kyle.millar@ieee.org</p> END:VEVENT END:VCALENDAR