Source searching is an important task in many real world applications and it has been widely studied using various types of robotic systems. In this task, the robot has to first detect the presence of the source in the search space before it can continue to locate the source to its approximate location. In order to efficiently detect the source, the robot must optimize search space exploration. Lévy flight is one of the optimal algorithms for search space exploration. However, for a destructive type of source searching, Lévy flight is susceptible to two problems. Firstly, the possibility of immediate returns to the previously visited site is high. Secondly, robot tends to temporarily stuck at the boundary of the search space until a new angle that bring robot away from the boundary is generated. In this paper, the solutions to the two problems are proposed by imposing angle constraints to Lévy flight and by using boundary reflection, reflectively. The results show that the exploration capability of the Lévy flight with the proposed strategy is improved compared to the conventional Lévy flight model.