Drive-data-eval

Data Evaluation Client

This script fetches telemetry data from the Kuksa Data Broker, performs anomaly detection using a pre-trained LSTM/GRU TensorFlow model, and publishes the results to a new topic in the data broker. Optionally, it logs telemetry data to a JSON file for further analysis.

---

Setup

Hardware Requirements

• Raspberry Pi or any system capable of running Python.
• Access to a Kuksa Data Broker instance.

Software Requirements

• Python 3.7+
• TensorFlow 2.x
• kuksa-client Python library
• Standard Python libraries: numpy, json, argparse

Install required Python libraries:

pip install tensorflow numpy

---

Usage

Run the Script

Save the script as data_eval_client.py and execute it with the following command:

python3 data_eval_client.py -th <host> -tp <port> -m <model_path> [-w <write_to_file>] [-o <output_file>] [-i <interval>] [--verbose]

Arguments

Argument	Description	Default
-th, --host	Hostname of the Kuksa Data Broker.	localhost
-tp, --port	Port of the Kuksa Data Broker.	55555
-w, --write	Whether to write telemetry data to a JSON file.	False
-o, --output	Output JSON file path for telemetry data.	telemetry_data.json
-m, --model	Path to the pre-trained TensorFlow model. Required.
-i, --interval	Time interval (in seconds) between data polling.	0.1
--verbose	Print telemetry data and predictions to the console.	Disabled

---

Example Command

python3 data_eval_client.py -th localhost -tp 55555 -m lstm_model.h5 -w True -o telemetry.json -i 0.5 --verbose

---

Troubleshooting

1. Connection Issues:

   • Verify that the Kuksa Data Broker is running and accessible at the specified host and port.
   • Check the network connectivity between the client and the broker.

2. TensorFlow Model Compatibility:

   • Ensure the LSTM/GRU TensorFlow model matches the input shape and features used in the script.

3. Error Handling:

   • Enable verbose mode (--verbose) to debug issues with telemetry data or predictions.

---

For further questions or support, feel free to reach out!