Binary Exponential Back Off for Tabu Tenure in Hyperheuristics

In this paper we propose a new tabu search hyperheuristic which makes individual low level heuristics tabu dynamically using an analogy with the Binary Exponential Back Off (BEBO) method used in network communication. We compare this method to a reduced Variable Neighbourhood Search (rVNS), greedy and random hyperheuristic approaches and other tabu search based heuristics for a complex real world workforce scheduling problem. Parallelisation is used to perform nearly 155 CPU-days of experiments. The results show that the new methods can produce results fitter than rVNS methods and within 99% of the fitness of those produced by a highly CPU-intensive greedy hyperheuristic in a fraction of the time.

• Type: Conference
• Year: 2009
• Where: EVOW
• Authors: Stephen Remde, Keshav P. Dahal, Peter I. Cowling, Nic Colledge
• Keywords: Artificial Intelligence, EVOW 2009, Low Level Heuristics, Search Based Heuristics, Tabu Search