This page was generated from docs/devices/xylo-imu/imu-if.ipynb. Interactive online version: Binder badge

∿ The Xylo™ IMU preprocessing interface

Xylo™ IMU contains a set of preprocessing blocks designed for encoding IMU data as events for inference using the SNN core on Xylo. This notebook explains the design and operating principles of the IMU interface, and shows how to use the IMU IF simulation.

[10]:
# - Imports and configuration
import numpy as np

# - Plotting and config
import matplotlib.pyplot as plt

plt.rcParams["figure.figsize"] = [9.6, 3.6]
plt.rcParams["figure.dpi"] = 1200
plt.rcParams["font.size"] = 12

try:
    from rich import print
except:
    pass

from IPython.display import Image

import warnings

warnings.filterwarnings("ignore")

Block diagram of IMU interface

[11]:
Image('IMU-IF-block-level.png')
[11]:
../../_images/devices_xylo-imu_imu-if_3_0.png

Raw IMU data is obtained at 200 Hz from the IMU sensor, providing quantized three-channel (x, y, z) samples.

Depending on the applicaiton use-case, an IMU sensor might have an arbitrary or changing orientation in the real world. The IMU interface contains a set of algorithms for estimating and correcting for sensor orientation, such that the gravitity vector is always aligned with the -Z axis. This rotation correction module can be bypassed.

The IMU data is analysed by a tunable bank of bandpass filters, which can separate the data frequencies with central frequencies between 0..20Hz. Up to five filters can be used for each (x, y, z) channel of the IMU data.

Two alternative strategies are available for event encoding. The first quantizes and scales the filtered data to a maximum of 15 events per time-step per channel. The second uses a digital integrate-and-fire neuron to encode events.

A full simulation of the IMU encoding interface is available in Rockpool, via the IMUIFSim module and the devices.xylo.syns63300.imuif package.

[12]:
# - Import Rockpool modules
from rockpool.devices.xylo.syns63300 import Quantizer, IMUIFSim
from samna.xyloImu.configuration import InputInterfaceConfig

[13]:
Image('imu-ifsim-module.png')
[13]:
../../_images/devices_xylo-imu_imu-if_7_0.png

The high-level module IMUIFSim encapsulates the entire preprocessing toolchain, and permits configuration of the encoding approach. Internally the several blocks in the encoding chain also have corresponding Rockpool modules (compare the two figures above). It is also possible to access the simulations of individual blocks, using the Rockpool modules RotationRemoval, FilterBank, ScaleSpikeEncoder and IAFSpikeEncoder.

The IMU IF simulation expects integer-quantized data with 16 bits per sample per channel (i.e. -32’768..32’767). This can conveniently be simulated using the Quantizer module.

[14]:
## Load and quantize an IMU sample record
with open("data.npy", "rb") as f:
    data = np.load(f)

quantizer = Quantizer(scale=0.49, num_bits=16)

data_quantized, _, _ = quantizer(data)
data_quantized.shape

plt.figure()
plt.plot(data_quantized[0])
plt.title(f"Quantized IMU Sample Record")
plt.show()

../../_images/devices_xylo-imu_imu-if_9_0.png

Using the high-level simulation interface

The Rockpool module IMUIFSim uses default configuration parameters covering a reasonable set of use cases. By default rotation removal is switched on, and the ScaleSpikeEncoder is used.

You create and interact with the IMU IF simulation module identically to any other Rockpool module:

[15]:
# - Generate an IMU IF simulation with default parameters
mod_IMUIF = IMUIFSim()
print(mod_IMUIF)
IMUIFSim  with shape (3, 15) {
    ModSequential 'model' with shape (3, 15) {
        RotationRemoval '0_RotationRemoval' with shape (3, 3) {
            ModSequential 'sub_estimate' with shape (3, 9) {
                SubSpace '0_SubSpace' with shape (3, 9)
                SampleAndHold '1_SampleAndHold' with shape (9, 9)
            }
        }
        FilterBank '1_FilterBank' with shape (3, 15)
        ScaleSpikeEncoder '2_ScaleSpikeEncoder' with shape (15, 15)
    }
}
[16]:
# - Pass an IMU sample through the simulation
result, _, _ = mod_IMUIF(data_quantized)

# - Display the encoded result
plt.figure()
plt.imshow(result[0].astype('float').T, aspect='auto', origin='lower')
plt.title(f"Preprocessed IMU Sample Record")
plt.show()
../../_images/devices_xylo-imu_imu-if_13_0.png

Export the hardware configuration, for use in configuring a Xylo IMU chip, using the export_config() method.

[17]:
config = mod_IMUIF.export_config()
print(config)
xyloImu::configuration::InputInterfaceConfig(enable=1, configuration_timeout=3.000000, estimator_k_setting=4,
select_iaf_output=0, bypass_jsvd=0, update_matrix_threshold=10, delay_threshold=500, bpf_bb_values={ 6 5 4 3 2 6 5
4 3 2 6 5 4 3 2 }, bpf_bwf_values={ 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 }, bpf_baf_values={ 9 10 11 12 13 9 10 11 12 13 9
10 11 12 13 }, bpf_a1_values={ -64458 -63288 -60684 -54529 -39246 -64458 -63288 -60684 -54529 -39246 -64458 -63288
-60684 -54529 -39246 }, bpf_a2_values={ 31754 30771 28888 25417 19378 31754 30771 28888 25417 19378 31754 30771
28888 25417 19378 }, scale_values={ 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 }, iaf_threshold_values={ 1024 1024 1024 1024
1024 1024 1024 1024 1024 1024 1024 1024 1024 1024 1024 })

Given a saved hardware configuration, you can also generate an IMU IF simulation based on that config, using the from_config() class method:

[18]:
mod = IMUIFSim.from_config(config)
print(mod)
IMUIFSim  with shape (3, 15) {
    ModSequential 'model' with shape (3, 15) {
        RotationRemoval '0_RotationRemoval' with shape (3, 3) {
            ModSequential 'sub_estimate' with shape (3, 9) {
                SubSpace '0_SubSpace' with shape (3, 9)
                SampleAndHold '1_SampleAndHold' with shape (9, 9)
            }
        }
        FilterBank '1_FilterBank' with shape (3, 15)
        ScaleSpikeEncoder '2_ScaleSpikeEncoder' with shape (15, 15)
    }
}

Next steps

See 🧑‍🏫 Specifying IMU preprocessing parameters for details of how to configure the preprocessing and encoding modules.