Tabu search algorithms are becoming very popular in operational research community. A lot of works and studies were carried out from the ÿrst presentation of Glover. The development of tabu search techniques concerns in almost all cases combinatorial problems, and we found very few papers about continuous problems. In this work, we brie y classify and describe the main continuous approaches to tabu search, then we will present a novel algorithm which explores a grid of points with a distance dynamically deÿned, it collapses to a local minimum then it continues the search from that point accepting some non-improving points to allow the exploration of new regions of the domain. The proposed algorithm is deterministic with a little random component triggered only when loop conditions are detected and it contains a simple vocabulary building mechanism and a diversiÿcation procedure. Finally we show some comparisons with other optimization algorithms and a possible application of this method to an engineering problem.