TV White Space Channel Allocation with Simulated Annealing as Meta Algorithm Bo Ye ESBE, London South Bank Uni. London SE1 0AA, UK Maziar Nekovee BT Research, Polaris 134 Adastral Park, Martlesham, Suffolk IP5 3RE, UK Anjum Pervez, and Mohammad Ghavami ESBE, London South Bank Uni. London SE1 0AA, UK Abstract—When developing a TV White Space (TVWS) system with the available TV spectrum after digital switchover, channel allocation for TVWS devices to avoid interference becomes one of the most challenging problems. In this paper, Simulated Annealing (SA) is applied to solve the problem to minimize the total interference. However, the performance of SA applied to the problem depends on the appropriate choice of several key parameters. Hence, all parameters are listed and experiments with all possible combinations are done and the parameters that perform best are chosen manually. Manual selection requires a lot of experiments and is not effective. Therefore, an algorithm using SA as meta algorithm is proposed to choose the parameters automatically for SA. Finally the result of parameters selected automatically is compared with the result of those selected manually. We will show that automatic selection performs better compared with manual selection. Index Terms—Simulated Annealing, Hyper-heuristics Algo- rithm, Channel Allocation, TV White Space. I. I NTRODUCTION The proposed switchover from analogue to Digital TV (DTV) in the UK and elsewhere in Europe and the United States has generated particular interest in cognitive wireless networks strategies to find new solutions for Radio Resource Management. After the switchover has been completed, a significant amount of RF spectrum within the existing TV band which are also known as TV White Space (WS) will become vacant for sharing. It is strongly believed that the devices with cognitive capability will be the prime contenders for the dynamic spectrum access to TVWS. When developing a TVWS system, three methods of sensing, beacons and Geo- location combined with database are used to avoid interference between TVWS devices and the primary TV band users. In sensing, TVWS devices detect the presence of TV signals and use the spectrum not used by the primary TV band users. In systems using beacons, devices transmit when they receive a beacon signal and the information of available channels. In our project, Geo-location is used to allocate available spectrum at a certain time in a given position. In Geo-location systems, all devices including TV transmitters and TVWS devices will register in a database and send the information of position, power and so on to the database. Furthermore, TVWS devices ideally should not cause inter- ference to their adjacent devices. The challenge is managing co-channel interference. For a large network with lots of devices, allocating the channels among all devices to minimize the total interference is an NP-complete problem. Briefly, this means that globally optimal solutions to these problems cannot be found because the computing time required to do so grows exponentially with the problem size. Hence, heuristic methods must be used, which find solutions that are close to optimal in a tractable amount of time. A distributed channel allocation was proposed in [1] based on Simulated Annealing (SA). Authors developed a channel allocation scheme to minimize the total interference of all Access Points (AP) in WLAN. In [1], authors proposed a dis- tributed SA to solve the dynamic channel allocation problem in 802.11b/g based HD-WLANs. When using SA, authors did not mention the temperature initialization, the details of choosing cooling schedule. And they did not give any other methods of generating neighbor state. Genetic Algorithm (GA) is a search heuristic that mimics the process of natural evolution. This heuristic is routinely used to generate useful solutions to optimization and search problems. GAs belong to the larger class of Evolutionary Algorithms (EA), which generates solutions to optimization problems using techniques inspired by natural evolution, such as inheritance, mutation, selection, and crossover. In [2, 3, 4, 5], GA is used to find a channel allocation from a search space to satisfy some constraints. In [6], authors proposed a fully distributed and self-managed algorithm. In [7], the channel assignment problem is solved by a neural network algorithm. In [8], authors proposed a hybrid algorithm that combines GA and SA. The algorithm will trap in local minima, because in the mutation step, authors allow the selected AP to choose the channel used least among its neighboring APs. With such a choice, the search strategy moves to the domain of local search, and the mutation used to jump out the local minima loses its original meaning. SA is a powerful technique for finding near-optimal solu- tions to NP-complete problem. However, SA relies on parame- ters that have to be fixed before the execution of the algorithm. It is usually very hard to choose appropriate parameters. A good choice of the parameters depends on the particular problem. Determining a good parameter setting often requires the execution of lots of time-consuming experiments with the aggravation that the best parameter setting for one problem is of limited utility for any other problem. In this paper, our contribution is to list all the parameters which involve our own approach, to list most of classic methods of the CROWNCOM 2012, June 18-20, Stockholm, Sweden Copyright © 2012 ICST DOI 10.4108/icst.crowncom.2012.248570