Converting multistatic radar measurements into Cartesian estimates of target position and velocity allows a linear Kalman filter, but the accuracy of each estimate depends on the transmitted waveform and on the geometry. Joint waveform and Cartesian estimate selection has so far been studied for one transmitter and without clutter. We extend it to several transmitters and to tracking in clutter. A new Cartesian estimate is formed from the bistatic ranges of two transmitters at one receiver, with a closed-form solution and an unscented-transform covariance. In clutter, candidate estimates formed from pairs of validated detections are combined by a probabilistic data association (PDA) filter, and the selection cost includes a clutter degradation factor computed by Monte Carlo integration. Simulations with Gaussian linear frequency-modulated (LFM) pulses show that a second transmitter reduces the position root mean square error (RMSE) by about 20% without clutter, mostly through its own channels; the new estimate adds about 1% in position and 5% in velocity. With the first-order covariance, 10.5% of the tracks are lost. In clutter, the reduction over the tracks kept is about 22% and 30% at the lowest and highest densities (200 runs) and 21% at the middle density (1000 runs), but when lost tracks are counted through an error cutoff it is significant only at the lowest density. The clutter factor reduces the position error by up to 30%. An unscented PDA filter on the raw measurements of the same two channels, chosen by the proposed selection, is 11% to 34% more accurate, which is the price of keeping a linear filter. A third transmitter adds little, and error maps show that the placement of the nodes matters strongly........................ Keywords:.................multistatic radar; adaptive waveform selection; Cartesian estimate; probabilistic data association; unscented transform; target tracking.