This project focuses on the implementation of optimized Linear and DNN regression models for inter-vehicle distance prediction in a Cooperative Adaptive Cruise Control (CACC) application. It leverages Tensorflow Lite to create optimized models through quantization and pruning for realtime inferencing on Raspberry Pi and On-board Unit (OBU) of Connected Autonomous Vehicles.
Major aspect of the project include:
- The cacc_application.ipynb file contains the different steps taken to load, clean and noromalize traffic data. This is followed by training, optimizing and conversion of a tensforlow saved model to tensorflow lite model. It finally describes how to export the final .tflite model.
- The predict.py script uses TFlite interpreter on the embedded device to predict inter-vehicle distance of the CACC application using six vehicle mobility features. It describes how to Load the trained TFLite model and how to allocate tensors.
This project leverages the following frameworks.
Access the cacc_application.ipynb file to train and optimize the TFlite model. Transfer the the TFlite models to a Raspberry Pi or OBU.
This is an example of how to list things you need to use the software and how to install them.
- Install Python 3.7.3 on the Raspberry Pi/OBU.
- Install TensorFlow 2.3.0 for Raspberry Pi3+/4.
- Generate mobility training data (with 6 features in this example) on ns-3.
-
Clone the repo
git clone https://github.com/simeonbabatunde/embedded-deep-learning-for-autonomous-vehicles
-
Train and export an optimized TFlite model using cacc_application.ipynb
-
Copy the TFlite model and predict.py into a directory on the Raspberry Pi/OBU.
-
Run predict.py to initiate inferencing on the edge device.
chmod+x predict.py ./predict.py
Distributed under the MIT License. See LICENSE for more information.
Simeon Babatunde - @simbabatunde - babatunde.simeon@gmail.com
Project Link: https://github.com/simeonbabatunde/embedded-deep-learning-for-autonomous-vehicles