Search Results

In this paper, an integrated approach that overcomes the tool and process inefficiencies in model-based development (MBD) of automotive embedded control unit (ECU) software is presented. This is done by first leveraging the Model-in-the-loop (MiL) technique on the desktop PC environment where the use of state-of-the-art plant models significantly accelerates the pace of software design, testing and calibration. Thereafter, the automatic generation of ECU code from the models allows the Software-in-the-Loop (SiL) technique to be used. This step brings a higher level of confidence in the implementation of the software without requiring ECU hardware. Finally, the software is downloaded to a development ECU and the Hardware-in-the-loop (HiL) technique is used to verify that the software behavior is identical to that observed in the MiL and SiL environments.

While model-based design improves several control algorithm validation steps, the AUTOSAR methodology describes a top-down configuration approach for series production of ECU networks. This paper shows how model-based control-algorithm validation can be interleaved with the AUTOSAR configuration approach. This comprises the integration of a control algorithm to atomic software components as well as a virtual integration of composition components. Last but not least, a survey of resource-analysis techniques for RTE-based ECUs is presented.

The process of developing new functions and software for powertrain embedded control units (ECUs) is undergoing a sea change. One of the reasons is the pressure to meet regulatory requirements (e.g. emissions, fuel economy) in addition to managing the normal growth in software complexity. Traditionally, the vehicle manufacturer (OEM) would write new specifications and hand them over to the software supplier (Tier1). This process required a substantial review and testing process at the OEM to ensure that the requirements were interpreted and implemented as indented. Today, a number of parties are involved in creating specifications (including the software supplier and 3rd party engineering companies), thus making the verification task even more complex.

The use of automated test tools for automotive ECU development has been increasing in recent years, in particular for software testing, system validation and verification, and regression testing. One of the challenges for the ECU development community is that test cases created at a specific phase of software development cannot be easily re-used upstream or downstream in the development process. For example, test cases developed for a model-in-the-loop (MiL) test environment can not be easily re-used on a hardware-in-the-loop (HiL) tests system. This results in significant costs associated with re-engineering test cases and/or poor software quality. At ETAS, we understand that a critical aspect of test case design is the ability to share and re-use test scripts on different HiL hardware platforms and across different development environments (e.g. MiL, SiL, and HiL).

Automotive controls engineers have traditionally used bypass rapid prototyping techniques to quickly try their new ideas before integrating them into the final embedded control unit (ECU) software. In order to make this possible, switches (or “hooks”) are required to be inserted in the ECU source code. The development disruption and costs associated with this can be extremely high. ETAS has developed a unique patent pending technology which solves this problem in a reliable way. This technology, to be productized under the name “EHOOKS”, allows the controls engineer to quickly add the necessary hooks in the base ECU software without any modifications to the source code. EHOOKS technology, therefore, enables “rapid” software prototyping and testing.

Over the last two decades, adoption of model-based techniques for the development of ECU software has resulted in major gains in productivity across the automotive industry. However, the fact remains that the majority of the ECU software today is still hand-written using the “C” programming language. Further, the need to shorten the development time, reduce costs and increase the quality of the ECU software has driven companies to adopt virtual (PC-based) simulation techniques rather than rely on expensive in-vehicle and dynamometer set-ups. This has lead to a situation where the two development philosophies (models and hand-written code) need to be properly integrated in order to fully capitalize on the advantages of PC-based techniques. For the complete ECU system to be simulated, typically, automatically generated C-code from other tools must be integrated as well.