## Efficient, Guaranteed Search with Multi-Agent Teams

Citations: | 7 - 3 self |

### BibTeX

@MISC{Hollinger_efficient,guaranteed,

author = {Geoffrey Hollinger and Sanjiv Singh and Athanasios Kehagias},

title = {Efficient, Guaranteed Search with Multi-Agent Teams},

year = {}

}

### OpenURL

### Abstract

Abstract — Here we present an anytime algorithm for clearing an environment using multiple searchers. Prior methods in the literature treat multi-agent search as either a worst-case problem (i.e., clear an environment of an adversarial evader with potentially infinite speed), or an average-case problem (i.e., minimize average capture time given a model of the target’s motion). We introduce an algorithm that combines finite-horizon planning with spanning tree traversal methods to generate plans that clear the environment of a worst-case adversarial target and have good average-case performance considering a target motion model. Our algorithm is scalable to large teams of searchers and yields theoretically bounded average-case performance. We have tested our proposed algorithm through a large number of experiments in simulation and with a team of robot and human searchers in an office building. Our combined search algorithm both clears the environment and reduces average capture times by up to 75 % when compared to a purely worst-case approach. I.

### Citations

417 | The Player/Stage project: Tools for multi-robot and distributed sensor systems
- Gerkey, Vaughan, et al.
- 2003
(Show Context)
Citation Context ... network, and the entire team is guided by a decentralized implementation of Algorithm 3. The robot’s position is determined by laser AMCL, and it is given waypoints through the Player/Stage software =-=[15]-=-. The humans input their position through the keyboard, and they are given waypoints through a GUI. In the first experiment, all three searchers (humans and robots) were assigned as G-GSST searchers. ... |

128 |
The complexity of searching a graph
- Meggido, Hakimi, et al.
- 1988
(Show Context)
Citation Context ...r of a graph as the number of searchers needed to guarantee the capture of an evader on that graph. Megiddo et al. later showed that finding the edge search number is NP-complete for arbitrary graphs =-=[5]-=-. Several variations of the edge search problem appear in the literature that place restrictions on the form of the cleared set: the nodes on the graph that have been cleared of a potential target. Co... |

86 | Visibility-based pursuit-evasion in a polygonal environment
- Guibas, Latombe, et al.
- 1997
(Show Context)
Citation Context ...lternative formulations of the guaranteed search problem. LaValle et al. formulated the search problem in polygonal environments and presented a complete algorithm for clearing with a single searcher =-=[6]-=-. However, their algorithm shows poor scalability to large teams and complex environments. Gerkey et al. demonstrate how a stochastic optimization algorithm (PARISH) can be used to coordinate multiple... |

52 |
Pursuit-evasion in a graph,” in Theory and Applications of Graphs
- Parsons
- 1978
(Show Context)
Citation Context ...WORK As mentioned above, literature in autonomous search can be partitioned into average-case and worst-case problems. The initial formulation of the worst-case graph search problem is due to Parsons =-=[4]-=-. He described a scenario where a team of agents searches for an omniscient, adversarial evader with unbounded speed in a cave-like topology represented by a graph. In this formulation, the edges of t... |

51 | Capture of an intruder by mobile agents
- Barrière, Flocchini, et al.
- 2002
(Show Context)
Citation Context ... wish either to locate a target in the environment or authoritatively say that no target exists. Such a scenario may occur in urban search and rescue [1], military operations, network decontamination =-=[2]-=-, or even aged care [3]. In some special cases, you may have a perfect model of how the target is moving; however, in most cases you will only have an approximate model or even no model at all. To com... |

43 | Efficient planning of informative paths for multiple robots
- Singh, Krause, et al.
- 2007
(Show Context)
Citation Context ...d presented a scalable heuristic [11]. Singh et al. discussed the related problem of multi-robot informative path planning and showed a scalable constant factor approximation algorithm in that domain =-=[12]-=-. In our prior work, we extended these approximation guarantees to average-case search with our Finite-Horizon Path Enumeration with Sequential Allocation (FHPE+SA) algorithm [13]. Our algorithm provi... |

41 | Robot and Sensor Networks for First Responders
- Kumar, Rus, et al.
(Show Context)
Citation Context ...g for a person, moving object, or contaminant. You wish either to locate a target in the environment or authoritatively say that no target exists. Such a scenario may occur in urban search and rescue =-=[1]-=-, military operations, network decontamination [2], or even aged care [3]. In some special cases, you may have a perfect model of how the target is moving; however, in most cases you will only have an... |

26 |
Searching and sweeping graphs: a brief survey
- Alspach
- 2004
(Show Context)
Citation Context ...r an environment of a worst-case target and improve average-case search performance based on additional 2 Note that several alternative versions of “node search” appear in the literature (see Alspach =-=[8]-=- for a survey). In one formulation, the evader resides in the edges of the graph, and these edges are cleared by trapping (i.e., two searchers occupy the adjacent nodes). In another, the pursuers have... |

22 | Selecting observations against adversarial objectives - Krause, McMahan, et al. |

19 | Parallel Stochastic Hill-Climbing with Small Teams
- Gerkey, Thrun, et al.
- 2005
(Show Context)
Citation Context ...y to large teams and complex environments. Gerkey et al. demonstrate how a stochastic optimization algorithm (PARISH) can be used to coordinate multiple robotic searchers in small indoor environments =-=[7]-=-. Though it is more scalable than complete algorithms, PARISH still requires considerable computation and 1 One exception would be a road network, which could be modeled as an edge search problem. The... |

14 | Planning under uncertainty for reliable health care robotics
- Roy, Gordon, et al.
- 2003
(Show Context)
Citation Context ...a target in the environment or authoritatively say that no target exists. Such a scenario may occur in urban search and rescue [1], military operations, network decontamination [2], or even aged care =-=[3]-=-. In some special cases, you may have a perfect model of how the target is moving; however, in most cases you will only have an approximate model or even no model at all. To complicate the situation f... |

14 | S.: A multi-robot strategy for rapidly searching a polygonal environment
- Sarmiento, Murrieta-Cid, et al.
(Show Context)
Citation Context ... of optimal solutions, researchers have proposed several approximation algorithms for average-case search. Sarmiento et al. examined the case of a stationary target and presented a scalable heuristic =-=[11]-=-. Singh et al. discussed the related problem of multi-robot informative path planning and showed a scalable constant factor approximation algorithm in that domain [12]. In our prior work, we extended ... |

13 | S.: Proofs and experiments in scalable, near-optimal search by multiple robots
- Hollinger, Singh
- 2008
(Show Context)
Citation Context ...rithm in that domain [12]. In our prior work, we extended these approximation guarantees to average-case search with our Finite-Horizon Path Enumeration with Sequential Allocation (FHPE+SA) algorithm =-=[13]-=-. Our algorithm provides near-optimal performance in our test domains, but it does not consider the possibility that the model is inaccurate. Thus, the search may last for infinite time if the target ... |

12 | Probabilistic planning for robotic exploration - Smith - 2007 |

7 | Anytime Guaranteed Search using Spanning Trees
- Kehagias, Ferguson, et al.
- 2008
(Show Context)
Citation Context ...represented as node search. 2 We proposed the Guaranteed Search with Spanning Trees (GSST) anytime algorithm that finds connected node search clearing schedules with a nearminimal number of searchers =-=[9]-=-. GSST is the first anytime algorithm for guaranteed search, and it is linearly scalable in the number of nodes in the graph, which makes it scalable to large teams and complex environments. Note that... |

3 |
Searching and sweeping graphs: a brief survey, Matematiche 59
- Alspach
- 2004
(Show Context)
Citation Context ...r an environment of a worst-case target and improve average-case search performance based on additional 2 Note that several alternative versions of “node search” appear in the literature (see Alspach =-=[8]-=- for a survey). In one formulation, the evader resides in the edges of the graph, and these edges are cleared by trapping (i.e., two searchers occupy the adjacent nodes). In another, the pursuers have... |

2 | Proofs and Experiments
- Hollinger, Singh
- 2008
(Show Context)
Citation Context ...rithm in that domain [12]. In our prior work, we extended these approximation guarantees to average-case search with our Finite-Horizon Path Enumeration with Sequential Allocation (FHPE+SA) algorithm =-=[13]-=-. Our algorithm provides near-optimal performance in our test domains, but it does not consider the possibility that the model is inaccurate. Thus, the search may last for infinite time if the target ... |