Search Results

Machine Learning (ML) based technologies are increasingly being used to fulfill safety-critical functions in autonomous and advanced driver assistance systems (ADAS). This change has been spurred by recent developments in ML and Artificial Intelligence techniques as well as rapid growth of computing power. However, demonstrating that ML-based systems achieve the necessary level of safety integrity remains a challenge. Current research and development work focused on establishing safe operation of ML-based systems presents individual techniques that might be used to gain confidence in these systems. As a result, there is minimal guidance for supporting a safety standard such as ISO 26262 - Road Vehicles - Functional Safety, to enable the development of ML-based systems. This paper presents a survey of recent ML literature to identify techniques and methods that can contribute to meeting ISO 26262 requirements.

Currently, Model Based Development (MBD) is the de-facto methodology in automotive industry. This has led to conversions of legacy code to Simulink models. Our previous work was related to implementing the C2M tool to automatically convert legacy code to Simulink models. While the tool has been implemented and deployed on few OEM pilot code-sets there were several improvement areas identified w.r.t. the generated models. One of the improvement areas identified was that the generated model used atomic blocks instead of abstracted blocks available in Simulink. E.g. the generated model used an ADD block and feedback loop to represent an integration operation instead of using an integrator block directly. This reduced the readability of the model even though the functionality was correct. Thus, as a user of the model, an engineer would like to see abstract blocks rather than atomic blocks.

Model-Based Development processes in the automotive industry typically use high-level modeling languages to build the reference models of embedded controllers. One can use formal verification tools to exhaustively verify these design models against their requirements, ensuring high quality models and a reduction in the cost and effort of functional testing. However, there is a gap, in terms of processes and tools, between the informal requirements and the formal specifications required by the verification tools. In this paper, we propose an approach that tries to bridge this gap by (i) identifying the verifiable requirements through a categorization process, (ii) providing a set of templates to easily express the verifiable requirements, and (iii) generating monitors that can be used as specifications in design verification tools. We demonstrate our approach using the Simulink Design Verifier tool for design verification of Simulink/Stateflow models.

The ever increasing number and complexity of software-controlled features in today's automotive vehicles mandate the use of numerous techniques and tools for ensuring the absence of any kind of defects in them. In this connection, one of the key areas of applied research today involves the identification, development and use of various analyses on Feature Requirements so that any defects can be caught early in the system development lifecycle.

The size and complexity of embedded software in automotive systems has been increasing rapidly. This makes the analysis of such systems difficult. For instance, in many analyses it is required to trace the dependences between variables in the software. E.g., in checking compliance to On-Board Diagnostics (OBD) standards one needs to ensure that only OBD compliant data-items are used (directly or indirectly) in an algorithm that is to be OBD compliant. Similarly, for safety analysis such as Design Failure Mode Effects Analysis (DFMEA), all the inputs to a safety critical system, all inputs to them, etc., have to be found, so that failure modes associated with these can be analysed. Currently such tracing of dependences is performed manually at great cost and effort. We describe the application of a technique (and tool) that automates the tracing of software dependence.

A key requirement for the development of safety-critical systems is the correctness of the tools used in their development process. Standards such as DO-178B mandate the qualification of tools used in the software engineering process of the systems to be certified at the highest levels of criticality. On the other hand, the increasing complexity of software requires the use of methodologies such as Model Based Development (MBD) that are highly tool intensive. MBD employs a suite of tools such as model-translators, code-generators, optimizers, simulators, etc., that can collectively be referred to as model-processors. A model-processor accepts a model in one language, and outputs a processed model in a possibly different language. Due to the increasing sophistication in modern modeling languages, model-processors are prone to implementation errors. Also, they are continuously evolving, resulting in differences in their behaviour across different releases.