TRULD - IoT Generative AI for Cargo

TRULD - IoT Generative AI for Cargo

Inspiration

In our interactions about ULD management with various stakeholders in the Air Cargo supply chain, we observe certain challenges regarding sustainability They include lack of standardization, limited ULD tracking systems, High maintenance and repair costs, and lack of data and metrics We also observe challenges in getting end to end visibility on a shipment or ULD as it goes through the various stages of the supply chain Our inspiration comes from this real-world business problem and trying to find a technological solution to the same based on ONE Record

What it does

Our solution: Uses Neural Networks to predict failures, enhance sustainability in ULD Management. Enables real-time tracking of ULDs by using IoT data Brings real-time visibility to the journey of Cargo.

How we built it

We used the following APIs Nexshore’s IATA ONE Record API Server and Database for storage and retrieval of Logistics Objects and IoT Data NE:ONE Server for information on Logistics objects OnAsset Server for real-time IoT Information We used React.js to create the front-end app and node.js for the backend API server We used Tensorflow for the Artificial Neural Networks modelling

Challenges we ran into

To be able to integrate machine learning models through Tensorflow libraries into node.js to implement Artificial Neural Networks based failure prediction

Accomplishments that we're proud of

We compute the risk of failure by feeding information from our IOT devices into an artificial neural network. We employ a sequential network model with two dense layers for this purpose that we built using the TensorFlow library. We also utilize a sigmoid function for training and train over a single data set 50 times in batches of 32 data points. The model is trained based on the data the user of the application generates, meaning that we can predict failure for various types of containers with a high degree of accuracy without actually having access to the data beforehand. This is to account for the variety of ULD containers in use with varying stress limits and tolerance to extreme conditions. We are proud of this development piece.

What we learned

That we can effectively utilize the IATA's data model with Interactive Cargo concepts, IoT and Machine learning to solve real world sustainability and visibility problems