Accelerating Automotive Code Migration with Generative AI

In the automotive industry, developers creating software for microcontrollers often face the task of converting a codebase from one microcontroller to another. The primary challenge is the tight coupling of the codebase to the current microcontroller, due to either performance requirements or its interaction with the microcontroller's peripherals. In this blog post, I explore how Generative AI can assist developers in navigating large codebases by providing intelligent code retrieval and understanding, using a real-world example from the Infineon AURIX TC4 microcontroller as a target device.The goal of this blog post is to demonstrate the potential of AI and transformer models for code migration. The examples provided are intended to showcase the methodology and highlight how these technologies can assist in navigating complex codebases. However, the implementation outlined here is a proof of concept, meant to illustrate the approach rather than serve as a complete, production-ready solution for developers.

The Use Case: Migrating to Infineon AURIX TC4

One of the common challenges in the automotive industry is migrating an existing codebase from one microcontroller to another. As a running example, I use the Infineon AURIX TC4 microcontroller. The process of converting a microcontroller codebase involves understanding how different peripherals, such as CAN controllers, are initialized and configured in the existing code, and then adapting this configuration to the new AURIX architecture.The complexity of such a migration becomes clear when dealing with extensive legacy codebases. Developers must scan and understand thousands of lines of code in the source code to be able to do the migration. In addition, they need to understand how certain functionality is implemented for the target microcontroller. This is where a Generative AI**-**based code retrieval and understanding tool comes into play. Such tools can help to gain a deeper understanding faster of the existing codebase as well as the target microcontroller. In this blog post, I show how the huge source of examples present at GitHub can be used as a data source for a Generative AI system to help a developer gain a better understanding, faster.

The Challenge: Navigating Large Codebases

Infineon provides a huge codebase with examples for the AURIX. When working with extensive codebases such as the one for the AURIX, developers often need to quickly locate and comprehend specific code sections. For instance, configuring and initializing CAN (Controller Area Network) communication is a common task in automotive software development. However, manually searching through thousands of lines of code to find relevant configuration functions can be time-consuming and error-prone.To address this, I set out to create an intelligent system that could:

Retrieve relevant code snippets based on a natural language query.
Provide explanations or further context about the retrieved code, helping developers understand its purpose and functionality.
Assist in migrating codebases by identifying and adapting key code segments for the AURIX TC4 microcontroller.

Solution Overview: Retrieval Augmented Generation Using Amazon Bedrock Knowledge Bases

The solution leverages a feature of Amazon Bedrock called Knowledge Bases. A Knowledge Base is a service providing out-of-the-box functionality to create embeddings for a set of attached data sources, as well as all functionality for answering questions using embeddings and a large language model. With that, the solution has the ability to summarize code which might be hosted in private repositories, which is particularly useful for adapting code to new architectures, such as the Infineon AURIX TC4 microcontroller.

Use of Amazon SageMaker and Amazon Bedrock

In the context of developing and deploying AI-based solutions, Amazon SageMaker and Amazon Bedrock play complementary roles. While SageMaker provides the infrastructure for managing the development environment, specifically handling Jupyter notebook instances, Bedrock offers a platform for interacting with various state-of-the-art generative AI models and additional functionality like Knowledge Bases and Agents.Integrating these two services, I created a Jupyter Notebook on Amazon SageMaker interacting with an Amazon Bedrock Knowledge Base. I would like to stress that the code presented below is self-contained. By combining the code blocks in one Jupyter Notebook, you are able to recreate my setup. Be aware that you need to create your own Amazon Bedrock Knowledge Base and insert the respective Knowledge Base ID in the code below. Please refer to the respective documentation describing how to set up an Amazon Bedrock Knowledge Base.The Knowledge Base I created contains a part of the public code repository containing code examples for the Infineon AURIX TC4 microcontroller.

Interaction: Amazon Bedrock and Amazon SageMaker

Here's a Step-by-Step Breakdown of My Approach:

Setup

This code installs required packages and imports modules to set up an environment for interacting with AWS Bedrock services. It configures AWS client settings with custom timeouts and disables retries, then initializes clients for 'bedrock-runtime' and 'bedrock-agent-runtime'. Finally, it starts an AWS session and retrieves the region name, preparing the script to perform operations with AWS Bedrock.

Querying the Knowledge Base

The following code invokes the retrieve_and_generate API using a specified Knowledge Base ID (kb_id). Please replace the dummy Knowledge Base ID with your actual one. It defines three example queries related to the configuration and usage of the VADC (Versatile Analog-to-Digital Converter) module of the Infineon AURIX microcontroller. For each query, it calls the retrieve_and_generate function to obtain a response and associated citations. The script then prints the query and response, and iterates through the citations to extract and display the URIs of retrieved references from a nested JSON structure. This demonstrates how to interact with the API to retrieve information and references from the knowledge base for specific technical questions.By leveraging this functionality, developers can ask questions about specific information and code examples related to their codebase. It allows them to retrieve answers to their questions and gain a deeper understanding of complex modules without manually sifting through extensive documentation or code, accelerating the development process.The three queries below show how a developer can approach a specific question. They might start with a general question in the area where the answer is expected. With the first answer, the question can be further specified. As a final question, the developer might ask for a code sample.

Analysis of the Answers

Below are the queries to the Knowledge Base and the respective answers. The generated answers correctly address the questions, providing accurate technical information relevant to the VADC module. The answers are well-structured, showing developers instructions on how to work with the Analog-to-Digital Converter of the Infineon AURIX microcontroller. By combining descriptive steps with actual code examples, the responses cater to both conceptual understanding and practical implementation needs. The inclusion of citations directs users to resources where they can find more detailed information or additional context.

Query: What is the resolution of an input channel of the Analog-to-Digital Converter (VADC)?Response: The Versatile Analog-to-Digital Converter (VADC) module of the AURIX TC27x microcontrollers has input channels with a resolution of up to 12 bits.Citations:

s3://micro-controller-migration-dschlei/Infineon-AURIX_ADC_Background_Scan_1_KIT_TC275_LK-TR-Training-v01_00-EN.pdf

Query: How do I configure the VADC module?Response: To configure the VADC module, you need to follow these steps:

Initialize an instance of the IfxVadc_Adc_Config structure with default values using the IfxVadc_Adc_initModuleConfig() function.
Optionally modify the configuration parameters in the IfxVadc_Adc_Config structure as needed, such as setting the digital and analog frequencies, enabling startup calibration, or selecting the low supply voltage.
Apply the configuration to the VADC module using the IfxVadc_Adc_initModule() function, passing the VADC module handle and the configured IfxVadc_Adc_Config structure.

Citations:

s3://micro-controller-migration-dschlei/code_examples/ADC_Background_Scan_1_KIT_TC275_LK/Libraries/iLLD/TC27D/Tricore/Vadc/Adc/IfxVadc_Adc.h
s3://micro-controller-migration-dschlei/Infineon-AURIX_ADC_Background_Scan_1_KIT_TC275_LK-TR-Training-v01_00-EN.pdf

Query: Please show me a code example for the configuration of the VADC module.Response: To configure the VADC module, you can follow these steps:

Create a configuration structure IfxVadc_Adc_Config and initialize it with IfxVadc_Adc_initModuleConfig().
Initialize the VADC module with IfxVadc_Adc_initModule() using the configuration from step 1.
Create a group configuration structure IfxVadc_Adc_GroupConfig and initialize it with IfxVadc_Adc_initGroupConfig().
Configure the group settings like group ID, trigger mode, arbitration, etc., in the group configuration structure.
Initialize the VADC group with IfxVadc_Adc_initGroup() using the group configuration from step 4.

Here is a coding example from the search results:

Citations:

s3://micro-controller-migration-dschlei/code_examples/ADC_Background_Scan_1_KIT_TC275_LK/Libraries/iLLD/TC27D/Tricore/Vadc/Adc/IfxVadc_Adc.h

Conclusion

The Generative AI-based approach for code retrieval and understanding is a promising tool for developers dealing with large and complex codebases. In the specific use case of migrating a codebase to the Infineon AURIX TC4 microcontroller, this system can significantly streamline the process by quickly retrieving relevant code sections and providing context or explanations to aid in the migration.Although challenges remain, particularly in processing and interpreting specialized code, this solution demonstrates how AI can assist in navigating and understanding automotive software.This project opens the door to further enhancements, such as incorporating more sophisticated models or expanding the system to handle multiple files and more complex queries. As AI continues to evolve, its role in software development, particularly in specialized domains like automotive systems, will only become more significant.

Any opinions in this post are those of the individual author and may not reflect the opinions of AWS.

Site Terms, Privacy, and more.