Issue with apply_doppler ?

[jj6142]

Hello,

I have been playing with Sionna ray tracer and tried to follow this guide:
https://nvlabs.github.io/sionna/examples/Sionna_Ray_Tracing_Introduction.html
The code there works fine as is and let's have it as starting point.

First, I try to add two more receivers by adding these lines to cell 6:
rx2 = Receiver(name="rx2",
position=[25,90,1.5],
orientation=[0,0,0])
scene.add(rx2)
rx3 = Receiver(name="rx3",
position=[35,70,1.5],
orientation=[0,0,0])
scene.add(rx3)
This is still fine. But when I try to modify apply_doppler such that each RX has their own velocity by modifying apply_doppler call in cell 13 like this:

rx_vel=tf.convert_to_tensor([[0,0,0],[1,0,1],[0,1,2]])
rx_vel=rx_vel[tf.newaxis,...]
print('rx_vel',rx_vel.shape) # this gives (1, 3, 3)

paths.apply_doppler(sampling_frequency=subcarrier_spacing,
num_time_steps=14, tx_velocities=[0,0,0], rx_velocities=rx_vel)

I will get the following error at the end. The dimension of rx_vel should be correct (I assume that dimension for rx_velocities should be [batch_size, num_rx, 3] and not [batch_size, num_tx, 3] as in the documentation.

Do I miss something?

---

InvalidArgumentError Traceback (most recent call last)
Cell In[27], line 10
8 print('rx_vel',rx_vel.shape)
9 # Apply Doppler shifts
---> 10 paths.apply_doppler(sampling_frequency=subcarrier_spacing, # Set to 15e3 Hz
11 num_time_steps=14, # Number of OFDM symbols
12 tx_velocities=[0,0,0], # We can set additional tx speeds
13 rx_velocities=rx_vel) # Or rx speeds
15 print("Shape of a after applying Doppler shifts: ", paths.a.shape)
17 a, tau = paths.cir()

File ~/python/miniconda3/envs/tf213/lib/python3.9/site-packages/sionna/rt/paths.py:427, in Paths.apply_doppler(self, sampling_frequency, num_time_steps, tx_velocities, rx_velocities)
422 # Compute the Doppler shift
423 # [batch_dim, num_rx, num_rx_ant, num_tx, num_tx_ant, max_num_paths]
424 # or
425 # [batch_dim, num_rx, 1, num_tx, 1, max_num_paths]
426 tx_ds = two_pidot(tx_velocities, k_t)/self._scene.wavelength
--> 427 rx_ds = two_pidot(rx_velocities, k_r)/self._scene.wavelength
428 ds = tx_ds + rx_ds
429 # Expand for the time sample dimension
430 # [batch_dim, num_rx, num_rx_ant, num_tx, num_tx_ant, max_num_paths, 1]
431 # or
432 # [batch_dim, num_rx, 1, num_tx, 1, max_num_paths, 1]

File ~/python/miniconda3/envs/tf213/lib/python3.9/site-packages/sionna/rt/utils.py:247, in dot(u, v, keepdim)
224 def dot(u, v, keepdim=False):
225 r"""
226 Computes and the dot (or scalar) product between u and v
227
(...)
245 is set to False.
246 """
--> 247 res = tf.linalg.matvec(tf.expand_dims(u, -2), v)
248 if not keepdim:
249 res = tf.squeeze(res,axis=-1)

File ~/python/miniconda3/envs/tf213/lib/python3.9/site-packages/tensorflow/python/util/traceback_utils.py:153, in filter_traceback..error_handler(*args, **kwargs)
151 except Exception as e:
152 filtered_tb = _process_traceback_frames(e.traceback)
--> 153 raise e.with_traceback(filtered_tb) from None
154 finally:
155 del filtered_tb

File ~/python/miniconda3/envs/tf213/lib/python3.9/site-packages/tensorflow/python/framework/ops.py:6656, in raise_from_not_ok_status(e, name)
6654 def raise_from_not_ok_status(e, name):
6655 e.message += (" name: " + str(name if name is not None else ""))
-> 6656 raise core._status_to_exception(e) from None

InvalidArgumentError: {{function_node _wrapped__BatchMatMulV2_device/job:localhost/replica:0/task:0/device:CPU:0}} In[0] and In[1] must have compatible batch dimensions: [1,1,1,1,1,3,1,3] vs. [1,3,1,1,1,19,3,1] [Op:BatchMatMulV2] name:

These are with tensorflow 2.13 and sionna 0.15.1

[jj6142]

I noticed another error related to this example. If I specify another transmitter i.e. I have 2 TX+3RX case, I will get the following error with paths.cir():

---

UnimplementedError Traceback (most recent call last)
Cell In[20], line 18
9 # Apply Doppler shifts
10 # apply_doppler(paths,sampling_frequency=subcarrier_spacing, # Set to 15e3 Hz
11 # num_time_steps=14, # Number of OFDM symbols
12 # tx_velocities=[0,0,0], # We can set additional tx speeds
13 # rx_velocities=[0,0,0]) # Or rx speeds
16 print("Shape of a after applying Doppler shifts: ", paths.a.shape)
---> 18 a, tau = paths.cir()
19 print("Shape of tau: ", tau.shape)

File ~/python/miniconda3/envs/tf213/lib/python3.9/site-packages/sionna/rt/paths.py:577, in Paths.cir(self, los, reflection, diffraction, scattering)
574 tau_ = tf.expand_dims(tau_, -1)
575 phase = tf.complex(tf.zeros_like(tau_),
576 -2PIself.scene.frequency*tau)
--> 577 a = a*tf.exp(phase)
579 return a,tau

File ~/python/miniconda3/envs/tf213/lib/python3.9/site-packages/tensorflow/python/util/traceback_utils.py:153, in filter_traceback..error_handler(*args, **kwargs)
151 except Exception as e:
152 filtered_tb = _process_traceback_frames(e.traceback)
--> 153 raise e.with_traceback(filtered_tb) from None
154 finally:
155 del filtered_tb

File ~/python/miniconda3/envs/tf213/lib/python3.9/site-packages/tensorflow/python/framework/ops.py:6656, in raise_from_not_ok_status(e, name)
6654 def raise_from_not_ok_status(e, name):
6655 e.message += (" name: " + str(name if name is not None else ""))
-> 6656 raise core._status_to_exception(e) from None

UnimplementedError: {{function_node _wrapped__Mul_device/job:localhost/replica:0/task:0/device:CPU:0}} Broadcast between [1,3,2,2,16,19,14] and [1,3,1,2,1,19,1] is not supported yet. [Op:Mul] name:

[jhoydis]

Hi,

Thanks for reporting these issues. I was able to reproduce them. They will be fixed in the next release. In case that you do not want to wait for so long, I could let you know the changes that need to be applied to the source code (three lines).

[faycalaa]

Fixed by v0.16.