<!DOCTYPE html>
<html>
  <head>

    <meta http-equiv="content-type" content="text/html; charset=UTF-8">
  </head>
  <body>
    <p>reminder: Please come and see Arundhati's thesis defense
      happening now<br>
    </p>
    <div class="moz-forward-container"><br>
      <br>
      -------- Forwarded Message --------
      <table cellpadding="0" cellspacing="0" border="0"
        class="moz-email-headers-table">
        <tbody>
          <tr>
            <th valign="BASELINE" align="RIGHT" nowrap="nowrap">Subject:
            </th>
            <td>Reminder - Thesis Defense - January 17, 2025 - Arundhati
              Banerjee - Learning based approaches to practical
              challenges in multi-agent active search</td>
          </tr>
          <tr>
            <th valign="BASELINE" align="RIGHT" nowrap="nowrap">Date: </th>
            <td>Fri, 17 Jan 2025 12:23:22 -0500</td>
          </tr>
          <tr>
            <th valign="BASELINE" align="RIGHT" nowrap="nowrap">From: </th>
            <td>Diane L Stidle <a class="moz-txt-link-rfc2396E" href="mailto:stidle@andrew.cmu.edu"><stidle@andrew.cmu.edu></a></td>
          </tr>
          <tr>
            <th valign="BASELINE" align="RIGHT" nowrap="nowrap">To: </th>
            <td><a class="moz-txt-link-abbreviated" href="mailto:ml-seminar@cs.cmu.edu">ml-seminar@cs.cmu.edu</a> <a class="moz-txt-link-rfc2396E" href="mailto:ML-SEMINAR@CS.CMU.EDU"><ML-SEMINAR@CS.CMU.EDU></a>,
              <a class="moz-txt-link-abbreviated" href="mailto:yyue@caltech.edu">yyue@caltech.edu</a></td>
          </tr>
        </tbody>
      </table>
      <br>
      <br>
      <meta http-equiv="content-type" content="text/html; charset=UTF-8">
      <p><b><i>Thesis Defense</i></b></p>
      <p>Date: January 17, 2025<br>
        Time: 1:00pm (EST)<br>
        Place: GHC 4405 & Remote<br>
        PhD Candidate: Arundhati Banerjee</p>
      <p><b>Title: Learning based approaches to practical challenges in
          multi-agent active search</b></p>
      <div>Abstract:<br>
        <div>Interactive decision making is essential for the
          functioning of autonomous agents in both software and embodied
          applications. Typically, agents interact in a multi-agent
          environment with the goal of fulfilling individual or shared
          objectives. In this thesis, we study the multi-agent adaptive
          decision making problem in the framework of Multi-Agent Active
          Search (MAAS) with a focus on applications like search and
          rescue, wildlife patrolling or environment monitoring with
          multi-robot teams.  <br>
          <br>
          Multi-Agent active search involves a team of robots (agents)
          deciding <i>when </i>and <i>where</i> to gather information
          about their surroundings, conditioned on their past
          observations, in order to estimate the presence and position
          of different objects of interest (OOIs) or targets. Agents
          communicate with each other asynchronously, without relying on
          a central controller to coordinate the agents'
          interactions. Realistically, inter-agent communications may be
          unreliable, and robots in the wild have to deal with noisy
          observations and stochastic environment dynamics. Our setup
          formalizes MAAS with practical models of real-world sensing,
          noise, and communication constraints for aerial and ground
          robots. </div>
        <div><br>
        </div>
        <div>Part I of this thesis studies the benefits of non-myopic
          lookahead decision making in MAAS with Thompson sampling and
          Monte Carlo Tree Search. Additionally, we consider a
          multi-objective pareto-optimization setup for cost-aware
          active search, highlighting the challenges due to partial
          observability, decentralized multi-agent decision making, and
          computational complexity with combinatorial action search
          spaces. In Part II, we focus on the practical challenges due
          to observation noise and dynamic targets in multi-agent active
          search and tracking. Our proposed algorithms using Bayesian
          filtering in these settings empirically demonstrate the
          importance of uncertainty modeling for inference and decision
          making with noisy observations due to sensor errors or
          environment non-stationarity. Part III shifts focus to
          generative models for decision making, particularly the
          applicability of diffusion for lookahead MAAS with observation
          noise. In the final part, we discuss the broader applicability
          of these methods in the context of multi-agent decision making
          in robotics and other applications with similar real world
          constraints. </div>
        <div><br>
          <b>Thesis Committee</b>: <br>
        </div>
        <div><font color="#000000"><span style="font-family:Helvetica">Jeff
              Schneider (Chair)</span><br style="font-family:Helvetica">
            <span style="font-family:Helvetica">Geoff Gordon</span><br
              style="font-family:Helvetica">
            <span style="font-family:Helvetica">Barnabás Póczos</span><br
              style="font-family:Helvetica">
            <span style="font-family:Helvetica">Yisong Yue (Caltech) </span></font><br>
        </div>
        <div><font color="#000000"><span
              style="font-family:Helvetica;font-size:12px"><br>
            </span></font></div>
        <div>Link to the draft document:</div>
        <div><a
href="https://drive.google.com/drive/folders/1YsZgROFeltk4TUVYo-9ldoSlsEkTjaAB?usp=sharing"
            class="moz-txt-link-freetext" moz-do-not-send="true">https://drive.google.com/drive/folders/1YsZgROFeltk4TUVYo-9ldoSlsEkTjaAB?usp=sharing</a><br>
        </div>
        <div><br>
          Zoom meeting link:<br>
        </div>
        <a
href="https://cmu.zoom.us/j/2231085641?pwd=j7ZOUGOHWPnaCe6c2FuCW3Xb0q8cWb.1&omn=99132356799"
          moz-do-not-send="true">https://cmu.zoom.us/j/2231085641?pwd=j7ZOUGOHWPnaCe6c2FuCW3Xb0q8cWb.1&omn=99132356799</a><br>
      </div>
    </div>
  </body>
</html>