BEGIN:VCALENDAR VERSION:2.0 PRODID:-//wp-events-plugin.com//6.6.4.4//EN TZID:Asia/Jerusalem X-WR-TIMEZONE:Asia/Jerusalem BEGIN:VEVENT UID:0-800@aerospace.technion.ac.il DTSTART;TZID=Asia/Jerusalem:20240328T110000 DTEND;TZID=Asia/Jerusalem:20240328T120000 DTSTAMP:20240325T073643Z URL:https://aerospace.technion.ac.il/events/simplification-for-efficient-d ecision-making-under-uncertainty-with-general-distributions/ SUMMARY:Simplification for Efficient Decision Making Under Uncertainty with General Distributions DESCRIPTION:Lecturer:Andrey Zhitnikov \n Faculty:TASP\n Institute:Technion – Israel Institute of Technology\n Location:Zoom\n Zoom: https://eur01.s afelinks.protection.outlook.com/?url=https%3A%2F%2Ftechnion.zoom.us%2Fj%2F 92401602138%3Fpwd%3DcUt5SVlkWHJoQW5IMFJ5ejJIRlFmZz09&\;data=05%7C02%7Ca eug-assistant%40technion.ac.il%7C782343b3aa0549112d6608dc40ddfaa8%7Cf1502c 4cee2e411c9715c855f6753b84%7C1%7C0%7C638456571053068967%7CUnknown%7CTWFpbG Zsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7 C0%7C%7C%7C&\;sdata=5GyPEem4R49tAIVVMb2rJ2%2FEC5YR8LjiZgndCpBLcUU%3D&am p\;reserved=0\n Abstract: \n\n\n\nPlanning under uncertainty in partially observable domains\, also known as Belief Space Planning (BSP)\, is an ess ential scientific inquiry to create Artificial Intelligence and a recurren t aspect in many real-life scenarios. The prevailing way to formulate such a decision making is the Partially Observable Markov Decision Process (PO MDP). The POMDP is notoriously hard to solve due to the multitude of possi ble future robot actions and observations along the planning horizon.\nIn this research\, we address the problem of efficient online BSP in continuo us domains. Towards this end\, we first introduce the adaptive multilevel simplification paradigm\, which aims to accelerate decision-making while p roviding performance guarantees. We then suggest a novel Probabilistic POM DP formulation that takes into account the variability of the POMDP elemen ts stemming from nonparametric representations. Based on this extension we formulate stochastic bounds over the cumulative robot reward and perform risk-aware BSP efficiently considering state-dependent rewards and informa tion-theoretic rewards\, such as differential entropy. In this work\, we s uggest an adaptive scheme to evaluate the theoretical differential entropy over a general belief surface. Finally\, we focus on Constrained POMDP. H ere\, we propose a novel Probabilistically Constrained belief-dependent PO MDP. Our constraint operator is belief-dependent for the first time to the best of our knowledge. It can be information related in the context of ex ploration or motivated by safety aspects. In both settings\, we accelerate online planning while providing guarantees on the impact of the accelerat ion.\n\n\n\n\n Details: \n CATEGORIES:Seminars LOCATION:Zoom END:VEVENT BEGIN:VTIMEZONE TZID:Asia/Jerusalem X-LIC-LOCATION:Asia/Jerusalem BEGIN:STANDARD DTSTART:20231029T010000 TZOFFSETFROM:+0300 TZOFFSETTO:+0200 TZNAME:IST END:STANDARD END:VTIMEZONE END:VCALENDAR