This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

AWR1443: Angle estimation without doppler FFT

Part Number: AWR1443

Hi,

I have been watching TI mmWave videos and one thing which I want to be clear of is, is it possible to calculate the angle of the object in front of radar without going into doppler FFT.

(my understanding is Range-FFT --> Doppler-FFT --> Angle-FFT), but is possible to do ( Range-FFT --> Angle-FFT).

that is, can we get just Range and angle of the object, instead of range, velocity and, angle. 

Also, in mmWave lecture series, in module 1.5, there is a question asked towards the end of the video "how can we estimate the angle of 2 objects which are 'stationary' and at the same range?"

-Regards

Sid

  • You can avoid velocity estimation if you think your application has no benefit from estimating motion. Note the angle FFT is done on spatial antenna data not temporal data i.e rx array corresponding to same tx (or a particular combination of tx) transmission (data in the same chirp) whereas velocity estimation is on temporal data i.e each rx's sample (corresponding to same tx ant transmission) in time over different chirps are used for velocity estimation. In TDM-MIMO however, which helps increase angle resolution by letting you use numTx * numRx(perTx) antennas, you are creating a virtual rx array that represents groups of rx (corresponding to same chirp type) separated in time so you need to do motion estimation and compensation before doing angle calculation (hence need to estimate velocity), otherwise if there is motion, then your angle estimate will be in error. But if your scene is known to have no motion, then you can use TDM-MIMO also and perform angle estimation on the virtual rx array without doing doppler estimation (in other words, your doppler correction is 0). The out of box demo in the SDK is a case of TDM-MIMO where doppler compensation is done before angle estimation.

    I think unless your scene is known to be static, generally it may help to do velocity estimation. When you don't do velocity estimation, you cannot detect two objects that are present at the same range and angle but are at different velocities. It may seem this should be physically impossible (e.g two physical objects like cars will have already collided and close to 0 velocity in this state) but the range/angle resolution (we are not talking about accuracy here) grid is not infinitesimally small but is a grid whose cell size is determined by the finite range and angle resolution (the angle resolution is generally not that good and as you go further, the cell widens physically in the azimuth direction for example). So it may be physically possible (without collision) for two objects to be present in the same range/angle resolution cell (hence are detected as a single object) but arrived there at different velocities (and continuing their motion without collision) in which case the doppler processing will be able to separate them (if their velocities are separable as per the velocity resolution and their velocities are within the max velocity).

  • Hello Siddhartha,
    Yes, you can skip doppler FFT and directly jump to angle FFT. After getting a object in range FFT you need to find the angle of the object using information across all the RX channels.
    But in this case you may not able to distinguish b/w two objects which are at the same range and same angle which might get separated if you use doppler otherwise.

    Regards,
    Jitendra
  • Hi Piyush,

    Thank you very much for the detailed explaination behind the scene.
    My reason of skipping doppler FFT is to reduce calculation inside the chip (limited memory) and also getting higher data rate in static environment.

    I have one more question, if we have everything in our room static, but the EVM itself is on a moving object (lets say a robot), so in that case is doppler correction be helpful for angle estimation? or can we skip it to directly get angle estimation?

    Also, is their any TI based software which can directly jump from range to angle? (again goal is to reduce calculation inside the chip)

    Kindest regards,
    Sid
  • Hi Jitendra,
    Thank you very much for the reply,
    I have one more question regarding doppler FFT which I have posted in this thread, please find time to have a look into that.


    Kindest regards,
    Sid
  • Velocity is relative, so every statement that can be made of objects in the scene moving while sensor is stationary are true when scene is static but sensor is moving. So for example, you will need to estimate and compensate for the sensor motion relative to your static scene if you use TDM-MIMO scheme.

    I am not aware of any TI demo that does both angle and range but does not do doppler but other TIers serving this forum may know better, I will let them comment.