GSST: anytime guaranteed search

Abstract

We present Guaranteed Search with Spanning Trees (GSST), an anytime algorithm for multi-robot search. The problem is as follows: clear the environment of any adversarial target using the fewest number of searchers. This problem is NP-hard on arbitrary graphs but can be solved in linear-time on trees. Our algorithm generates spanning trees of a graphical representation of the environment to guide the search. At any time, spanning tree generation can be stopped yielding the best strategy so far. The resulting strategy can be modified online if additional information becomes available. Though GSST does not have performance guarantees after its first iteration, we prove that several variations will find an optimal solution given sufficient runtime. We test GSST in simulation and on a human-robot search team using a distributed implementation. GSST quickly generates clearing schedules with as few as 50% of the searchers used by competing algorithms.

Improving the Efficiency of Clearing with Multi-agent Teams

“…2 We proposed the Guaranteed Search with Spanning Trees (GSST) anytime algorithm that finds connected node search clearing schedules with few searchers (Hollinger et al, 2008(Hollinger et al, , 2010. GSST is linearly scalable in the number of nodes in the graph, which makes it applicable to large teams and complex environments.…”

“…The FHPE+SA algorithm (Hollinger et al, 2009b) is utilized to determine the schedules for the average-case searchers in the current paper's combined algorithm. The proposed G-GSST algorithm in the current paper is an extension of the GSST algorithm (Hollinger et al, 2010). Unlike GSST, G-GSST allows for the optimization of clearing time and the implicit use of guards.…”

Search and pursuit-evasion in mobile robotics

“…The robots utilize beacons in the indoor environment to localize themselves, and they execute a team strategy to capture a target moving faster than the searchers. Hollinger et al (2010a) implemented a building clearing algorithm using a human-robot search team on a single floor of an office building. Their approach utilizes the clearing schedule of an underlying spanning tree as a heuristic to generate clearing schedules with few searchers and fast clearing times.…”

“…In one formulation, the evader resides in the edges of the graph (hence, despite the name, this is really an edge search problem), and these edges are cleared by trapping (i.e., two searchers occupy the adjacent nodes). Hollinger et al (2010a) discussed the properties of adversarial search when the evader resides on the nodes, and they show its formal relationship to edge search. Any node search clearing strategy is also an edge search clearing strategy, but the opposite is not true.…”

Exact Solution for Search-and-Rescue Path Planning

“…For instance, dynamic programming [5], [7], [19], [20] or tree-based search techniques [21] may satisfactorily work under specific constraints and conditions but ultimately face the curse of dimensionality, showing poor scalability even for moderate size problem. This paved the way to the development of efficient heuristic and approximate methods.…”