High-speed odour sensing using miniaturised electronic nose
Data files
Oct 23, 2024 version files 14.68 GB
-
Dataset-FastMachineOlfaction.zip
14.68 GB
-
README.md
4.69 KB
Abstract
Animals have evolved to rapidly detect and recognise brief and intermittent encounters with odour packages, exhibiting recognition capabilities within milliseconds. Artificial olfaction has faced challenges in achieving comparable results-existing solutions are either slow; or bulky, expensive, and power-intensive-limiting applicability in real-world scenarios for mobile robotics. Here we introduce a miniaturised high-speed electronic nose; characterised by high-bandwidth sensor readouts, tightly controlled sensing parameters and powerful algorithms. The system is evaluated on a high-fidelity odour delivery benchmark. We showcase successful classification of tens-of-millisecond odour pulses, and demonstrate temporal pattern encoding of stimuli switching with up to 60 Hz. Those timescales are unprecedented in miniaturised low-power settings, and demonstrably exceed the performance observed in mice. For the first time, it is possible to match the temporal resolution of animal olfaction in robotic systems. This will allow for addressing challenges in environmental and industrial monitoring, security, neuroscience, and beyond.
https://doi.org/10.5061/dryad.pg4f4qrxz
The data comprises of recordings of complex odour stimuli using a high-speed a miniaturised electronic nose (e-nose). The odour stimuli were presented with a custom and high-fidelity olfactometer that was most prominently used in a recent mammalian olfaction study [1]. A photo-ionisation detector (PID) was used for ground-truth recordings and performance evaluations.
Details about the experimental protocol and the specificities of the used devices can be found in the manuscript [2].
Description of the data and file structure
This data repository consists of a nested directory structure (sensor modality - sensor mode - stimuli type - data files), containing a total of 4875 individual e-nose and 2670 PID recording trials. Each set of experiments is accompanied with an index.csv file, which specifies the trial names and experimental conditions, thus allowing for fast access of the trials-of-interest.
Index files
The index files are saved as comma-separated values (CSV), containing the following columns:
trial_idx: Index column, unique across all experiments
trial_id: Unique trial identifier combining 'trialidx', 'condition', 'kind', 'shape', 'concentration', 'gas1', 'gas2'
condition: Sensor mode. For the e-nose, this describes the sensor heater profiles.
block_idx: Index with respect to experiment start
kind: Type of stimulus. 'pulse': single gas odour pulse; 'corr': two correlated pulse trains; 'acorr': two anti-correlated pulse trains; 'plume': replayed odour plume (not used in the paper)
shape: Temporal structure of stimulus. 'pulse': pulse duration; 'corr': modulation frequency; 'acorr': modulation frequency; 'plume': mixed/co-located odour source or separated odour sources (not used in the paper)
concentration: Odour stimulus concentration, between 20% and 100%
gas1: Gas 1
gas2: Gas 2; empty for 'pulse'
t_stimulus: Timestamp of stimulus w.r.t. experiment initialisation, in the format 'd days hh:mm:ss'
n_saturated: Number of saturated sensors
E-nose trial files
The e-nose recording trials are saved as comma-separated values (CSV), containing the following columns:
index: Index column, unique across all experiments
timestamp: Timestamp from device boot
time_s: Time (s) with respect to odour onset
time_ms: Time (ms) with respect to odour onset
control_cycle_step_left: Heater power phase for sensor 1-4 (for cycles only)
control_cycle_step_right: Heater power phase for sensor 1-4 (for cycles only)
R_gas_1-R_gas_8: Gas sensor resistance values, in Ohm
T_heat_1-T_heat_8: Sensor heater temperature, in degree Celsius
p_mbar: Ambient pressure, in millibar
t_celsius: Ambient temperature, in degree Celsius
rh_percent: Ambient relative humidity, in percent
EB, IA, Eu, 2H, b1, b2, b_comp: Flow valve control values for the different odourants, values between 0.0 and 1.0
total: Sum of all flow valve control values, should add up to 1.0 at any time
PID trial files
The PID recording trials are saved as comma-separated values (CSV), containing the following columns:
index: Index column, unique across all experiments
time_ms: Time (ms) with respect to odour onset
timestamp: Timestamp from device boot
pid_V: PID output, in Volts
EB, IA, Eu, 2H, b1, b2, b_comp: Flow valve control values for the different odourants, values between 0.0 and 1.0
total: Sum of all flow valve control values, should add up to 1 at any time
Code/Software
For loading and displaying the data, please refer to the code in the following repository:
https://github.com/nkdnnlr/High-Speed-Odour-Sensing-Using-Miniaturised-Electronic-Nose
In particular, run the jupyter notebook called 00_display_data.ipynb, which will guide you through the necessary steps.
References
[1] Ackels, T., Erskine, A., Dasgupta, D., Marin, A.C., Warner, T.P., Tootoonian, S., Fukunaga, I., Harris, J.J. and Schaefer, A.T., 2021. Fast odour dynamics are encoded in the olfactory system and guide behaviour. Nature, *593 *(7860), pp.558-563.
[2] Dennler, N., Drix, D., Warner, T., Rastogi, S., Della Casa, C., Ackels, T., Schaefer, A.T., van Schaik, A. and Schmuker, M., 2024. High-speed odour sensing using miniaturised electronic nose. Science Advances (2024). preprint: arXiv:2406.01904.