Improvising Robot Patrol Routes
Tommaso Dreossi, Daniel J. Fremont, Sanjit Seshia

Citation
Tommaso Dreossi, Daniel J. Fremont, Sanjit Seshia. "Improvising Robot Patrol Routes". Talk or presentation, 26, October, 2016; Poster presented at the 2016 TerraSwarm Annual Meeting.

Abstract
Abstract: In classical robotic patrolling, a robot is required to visit several checkpoints at regular time intervals. However, a fixed, periodic schedule can be easily learned by adversaries, allowing them to evade the robot's surveillance. In this work we explore how to safely introduce unpredictability into the robot's behavior, recasting the patrolling problem in the framework of control improvisation. This formalization consists of 1) hard constraints that the robot must always satisfy (e.g. never hit obstacles), 2) soft constraints that can be relaxed (e.g. choose a path not much longer than the shortest path), and 3) a randomness requirement constraining the probabilities of picking paths (e.g. do not pick any single path too often). Designing a controller that simultaneously maximizes unpredictability and the rate at which checkpoints are visited can be difficult, and in fact impossible in some cases. We propose several methods to trade visit rate for randomness.

Electronic downloads

Internal. This publication has been marked by the author for TerraSwarm-only distribution, so electronic downloads are not available without logging in.

Citation formats

• HTML

  Tommaso Dreossi, Daniel J. Fremont, Sanjit Seshia. <a
  href="http://www.terraswarm.org/pubs/881.html"><i>Improvising
  Robot Patrol Routes</i></a>, Talk or
  presentation,  26, October, 2016; Poster presented at the
  <a
  href="http://terraswarm.org/conferences/16/annual"
  >2016 TerraSwarm Annual Meeting</a>.

• Plain text

  Tommaso Dreossi, Daniel J. Fremont, Sanjit Seshia.
  "Improvising Robot Patrol Routes". Talk or
  presentation,  26, October, 2016; Poster presented at the
  <a
  href="http://terraswarm.org/conferences/16/annual"
  >2016 TerraSwarm Annual Meeting</a>.

• BibTeX

  @presentation{DreossiFremontSeshia16_ImprovisingRobotPatrolRoutes,
      author = {Tommaso Dreossi and Daniel J. Fremont and Sanjit
                Seshia},
      title = {Improvising Robot Patrol Routes},
      day = {26},
      month = {October},
      year = {2016},
      note = {Poster presented at the <a
                href="http://terraswarm.org/conferences/16/annual"
                >2016 TerraSwarm Annual Meeting</a>},
      abstract = {Abstract: In classical robotic patrolling, a robot
                is required to visit several checkpoints at
                regular time intervals. However, a fixed, periodic
                schedule can be easily learned by adversaries,
                allowing them to evade the robot's surveillance.
                In this work we explore how to safely introduce
                unpredictability into the robot's behavior,
                recasting the patrolling problem in the framework
                of control improvisation. This formalization
                consists of 1) hard constraints that the robot
                must always satisfy (e.g. never hit obstacles), 2)
                soft constraints that can be relaxed (e.g. choose
                a path not much longer than the shortest path),
                and 3) a randomness requirement constraining the
                probabilities of picking paths (e.g. do not pick
                any single path too often). Designing a controller
                that simultaneously maximizes unpredictability and
                the rate at which checkpoints are visited can be
                difficult, and in fact impossible in some cases.
                We propose several methods to trade visit rate for
                randomness.},
      URL = {http://terraswarm.org/pubs/881.html}
  }
  

Posted by Tommaso Dreossi on 31 Oct 2016.
Groups: tools