Conquering The Traveling Salesman Problem

The basis of TSP is, given a list of cities and their distances, the task is to find a shortest possible tour that visits each city exactly once. The problem was first formulated in 1930 and is one of the most intensively studied problems in optimisation and used as a benchmark for many optimisation methods.

Now, as part of his final year project, Andrej Kazakov, who has just completed an MEng in Software Engineering at the University's School of Electronics and Computer Science (ECS) has come up with a new algorithm to solve TSP, which according to him "looks very promising".

Andrej, who was supervised by Dr Richard Watson at ECS' Science and Engineering of Natural Systems Group, reviewed existing solutions to TSP, the most popular approaches being local search and divide and conquer.

He then looked at a new algorithm called Building Block Hillclimber and applied a combination of approaches so that he substantially improved its performance and range of applications.

'I studied the algorithm to see how it performed in relation to a range of strategies,' Andrej said. 'The algorithm is still incomplete, but its current shortcomings have been thoroughly analysed and it has been substantially improved.'

Travelling Salesman Problem: Local Search and Divide and Conquer working together.