In this paper, we present an algorithm which lies in the domain of task allocation for a set of static autonomous radars with rotating antennas. It allows a set of radars to allocate in a fully decentralized way a set of active tracking tasks according to their location, considering that a target can be tracked by several radars, in order to improve accuracy with which the target is tracked. The allocation algorithm proceeds through a collaborative and fully decentralized auction protocol, using a collaborative auction protocol (Consensus Based Bundle Auction algorithm). Our algorithm is based on a double use of our allocation protocol among the radars. The latter begin by allocating targets, then launch a second round of allocation if they have resources left, in order to improve accuracy on targets already tracked. Our algorithm is also able to adapt to dynamism, i.e. to take into account the fact that the targets are moving and that the radar(s) most suitable for Tracking them changes as the mission progresses. To do this, the algorithm is restarted on a regular basis, to ensure that a bid made by a radar can decrease when the target moves away from it. Since our algorithm is based on collaborative auctions, it does not plan the following rounds, assuming that the targets are not predictable enough for this. Our algorithm is however based on radars capable of anticipating the positions of short-term targets, thanks to a Kalman filter. The algorithm will be illustrated based on a multi-radar tracking scenario where the radars, autonomous, must follow a set of targets in order to reduce the position uncertainty of the targets. Standby aspects will not be considered in this scenario. It is assumed that the radars can pick up targets in active pursuit, with an area of uncertainty corresponding to their distance.