Challenges of Continuous Monitoring, Improvement, and Deployment for Autonomous Driving Features: An Industrial Experience Report

Основные понятия

Implementing continuous monitoring, improvement, and deployment for autonomous driving features poses challenges in the automotive industry.

Аннотация

The article discusses challenges faced in implementing continuous development strategies for autonomous driving features. It highlights the need for safety adherence and documentation perspective required by industry standards such as ISO-26262 and ISO21448. The paper identifies challenges specific to the automotive domain that hinder the adoption of continuous monitoring, improvement, and deployment (CDDM) practices. These challenges are identified through interviews with domain experts and a literature study. Key points include the impact analysis before each iteration, safety-related challenges during field monitoring, continuous development, deployment, hardware limitations, software architecture design implications, verification and validation processes, and safety argumentation complexities. The study emphasizes the importance of exploring ways to support human intellectual work while utilizing automation to meet rapid development expectations.

Статистика

"ISO 26262:2018 (all parts), Road vehicles — Functional safety," standard, International Organization for Standardization, 2018.
"ISO/FDIS 21448, Road vehicles — Safety of the intended functionality," standard, International Organization for Standardization, 2022.
National Highway Traffic Safety Administration report on Safety Recall 22V-037 in 2022.

Цитаты

"As shown by the presented challenges...an overall conclusion about the challenges of introducing CDDM for safety-relevant products is to explore ways how to support the human in doing these intellectual activities."
"The main commonality in the challenges is the involvement of intellectual work that is required for every proposed or intended change of the product."
"CDDM shall not be seen only as a challenge but can also contribute towards a better engineering of safety-critical functions spanning from development phases at OEMs into actual operational phases when vehicles are in customers’ hands."

Ключевые выводы из

An Industrial Experience Report about Challenges from Continuous Monitoring, Improvement, and Deployment for Autonomous Driving Features

by Ali Nouri,Ch... в arxiv.org 03-15-2024

https://arxiv.org/pdf/2403.09474.pdf

Дополнительные вопросы

How can automation be effectively utilized to support human intellectual work in addressing challenges related to CDDM

Automation can play a crucial role in supporting human intellectual work when addressing challenges related to Continuous Development, Deployment, and Monitoring (CDDM) for autonomous driving features. Here are some ways automation can be effectively utilized:

Impact Analysis Automation: Implementing tools that can automate impact analysis before each iteration can significantly reduce the time and resources required for this critical step. These tools can analyze the effects of changes on various elements involved in providing the function and provide insights quickly.

Requirement Updates Automation: Automating the process of updating safety requirements (Functional Safety, Safety of Intended Functionality, Cyber-Security) when there is a change in design or functionality can ensure that all aspects are aligned without manual intervention.

Verification and Validation Automation: Utilizing automated testing frameworks and simulation tools can streamline verification processes by assessing if implemented elements meet their requirements efficiently. This automation reduces manual effort while ensuring thorough validation.

Safety Argumentation Tools: Developing software tools that assist in creating, maintaining, and expanding safety argumentation documentation using natural language processing capabilities could enhance traceability and understanding across different teams within an organization.

By leveraging automation in these key areas, organizations working on autonomous driving features can improve efficiency, accuracy, and consistency while tackling the challenges associated with CDDM effectively.

What are some potential drawbacks or limitations of implementing continuous monitoring strategies for autonomous driving features

While continuous monitoring strategies offer numerous benefits for enhancing safety-critical systems like autonomous driving features, there are potential drawbacks or limitations to consider:

Data Overload: Continuous monitoring generates vast amounts of data from vehicles operating in diverse conditions which may lead to information overload. Managing this volume of data effectively to extract actionable insights without overwhelming analysts could be challenging.

Real-time Decision Making: Relying solely on continuous monitoring for real-time decision-making introduces risks as immediate responses based on monitored data may not always align with comprehensive risk assessments conducted during development phases.

Resource Intensiveness: Maintaining continuous monitoring systems requires significant resources including infrastructure setup costs, data storage solutions, analytics tools implementation, skilled personnel training which might pose financial constraints for some organizations.

Privacy Concerns: Collecting extensive data through continuous monitoring raises privacy concerns regarding user information protection especially with over-the-air updates where personal vehicle usage details are transmitted back to manufacturers.

How can lessons learned from non-safety-critical domains like Google and Facebook be applied to enhance safety-critical system development practices

Lessons learned from non-safety-critical domains such as Google and Facebook's approach towards continuous improvement practices hold valuable insights that could be applied to enhance safety-critical system development practices like those seen in autonomous driving features:

Agile Methodologies Adoption:

Implementing agile methodologies borrowed from tech giants allows for quicker adaptation to changing requirements.

2 .Iterative Development:

Embracing iterative development cycles enables rapid feature enhancements while maintaining focus on safety considerations throughout each phase.
3 .User-Centric Design:

Prioritizing user feedback loops similar to how non-safety-critical applications gather customer behavior data helps tailor AD functions more closely according to end-user needs.
4 .Continuous Testing & Feedback Loop Implementation:

Establishing robust testing mechanisms combined with feedback loops ensures ongoing refinement based on real-world performance metrics gathered post-deployment.
5 .Cross-Functional Collaboration Enhancement
- Encouraging collaboration between hardware engineers ,software developers,and other stakeholders mirrors successful interdisciplinary teamwork models observed at tech companies leading to better integration efforts essential for safe ADS deployment
By incorporating these lessons into safety-critical system development practices specific challenges relatedto CDDM within automotive domaincan be addressed more effectively resultingin improved overall product qualityand enhanced innovationcapabilities