We present an algorithm that allows swarms of robots to navigate in environments containing unknown obstacles, moving towards and spreading along 2D shapes given by implicit functions. Basically, a gradient descent approach augmented with local obstacle avoidance is used to control the swarm. To deal with local minima regions, we use a coordination mechanism that reallocates some robots as “rescuers” and sends them to help other robots that may be trapped. The main objective of this paper is to analyze the performance of this algorithm in terms of its completion rate and communication requirements as the number of robots increases. For this, a series of simulations are presented and discussed.