In this paper, the problem of adaptively selecting radar waveforms from a pre-dened library of waveforms is
addressed from an information theoretic perspective. Typically, radars transmit specic waveforms periodically,
to obtain for example, the range and Doppler of a target. Although modern radars are capable of transmitting
dierent waveforms during each consecutive period of transmission, it is hitherto unclear as to how these
waveforms must be scheduled to best understand the dynamic radar scene. In this paper, a new information
theoretic metric - directed information - is employed for waveform scheduling, and is shown to incorporate
the past radar returns to eectively schedule waveforms. We formulate this waveform scheduling problem in a
Gaussian framework, derive the corresponding maximization problem, and illustrate several special cases.