1 / 8

A Game-Theoretic Approach to Determining Efficient Patrolling Strategies for Mobile Robots

A Game-Theoretic Approach to Determining Efficient Patrolling Strategies for Mobile Robots Francesco Amigoni, Nicola Gatti, Antonio Ippedico. Scenario. Summary of Contributions. Problem: determination of an efficient patrolling strategy for a mobile robot Idea:

libby
Download Presentation

A Game-Theoretic Approach to Determining Efficient Patrolling Strategies for Mobile Robots

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. A Game-Theoretic Approach to Determining Efficient Patrolling Strategies for Mobile Robots Francesco Amigoni, Nicola Gatti, Antonio Ippedico

  2. Scenario

  3. Summary of Contributions • Problem: determination of an efficient patrolling strategy for a mobile robot • Idea: • model the scenario as an extensive-form game played by the patroller and the intruder • solve the game to find the strategy for the patroller

  4. The Proposed Model:Assumptions • Time is discrete, players play in turns • Environment with n places • Patroller detects the presence of the intruder (captures the intruder) when it is in the patroller’s current place • Intruder knows the strategy of the patroller • Patroller’s actions: move from one place to another one (incurring in different costs), movements can be between any pairs of places • Intruder’s actions: wait or attempt to enter a place • Entering a place takes d turns • The game ends either when the intruder is captured or has entered a place • Players payoffs are defined according to values attributed to places, to costs for moving between places, and to rewards for capturing the intruder • Intruder can be of different types, each one with different values for places

  5. patroller’s action intruder’s action patroller’s action … … … … … … … … The Proposed Model:Extensive-form Game • The intruder knows the patroller’s strategy and the patroller knows it commitment-based strategy for the patroller • Finding an optimal solution is not easy, basically because the environment can dynamically change and because the game is infinite-horizon approximate solution

  6. patroller’s action intruder’s action patroller’s action … … … … … … … … Solving the Game:Finding a Patrolling Strategy • Greedy approach: we consider a slice of the extensive-form game as an independent strategic-form game • Solving each slice means finding the next optimal action for the patrolling robot • A slice can be solved by resorting to a multi-LP [Conitzer and Sandholm, EC 2006] or to a MILP [Paruchuri et al., AAMAS 2008] mathematical programming formulation • Solution: mixed strategy for the patrolling robot: {γ1,γ2,…,γn}

  7. Experimental Results • The approach scales reasonably well with the number n of places (using the multi-LP formulation) and with the number of intruder’s types (using the MILP formulation) • The approach can be applied to different environments • Linear environment • Ring and star environments • The approach adapts to dynamic changes in the environment 25 nodes, 10 runs, 500 steps each

  8. Conclusions • We proposed a game-theoretic approach to determining strategies for patrolling robots • Modeling a patrolling situation as an extensive-form game • Finding an (approximate) solution of the game • Patrolling strategies found with our approach are efficient • Ongoing work • Optimal solutions for the extensive-form game • More realistic scenarios • Implementation on real robots

More Related