Dr. Manaswini Rath, VP and Global Head, Autonomous Driving, KPIT presented at the webinar with Automotive World. Here is a quick perspective and excerpts from the webinar.

Reading Time: 4 min


Dr Manaswini Rath, Vice President and Global Head, Autonomous Driving, KPIT

One of the key challenges faced in ADAS and AD development is Verification and Validation. Given the safety- critical nature of ADAS/AD, it’s important to ensure high levels of accuracy for this. This is where Virtual Simulation for Validation comes into play.

In her presentation, KPIT’s Dr Rath provides insights on how, by using virtual simulation, one can:
  • ensure maximum validation coverage
  • develop strategies to collect data and create the right scenarios
  • understand the KPIs against which to conduct tests
  • account for the Safety and SoTIF (Safety of The Intended Function) scenarios
  • and integrate with conventional requirement-driven verification
Achieving Maximum ADAS and AD feature-validation through data-driven virtual simulation methods

Data is key because it simulates the scenarios for tests. With KPITs unique KPI selector, the process of validating features and sensors moves up a notch and enhances their integrated approach to verification and validation for ADAS and AD, which includes:

  • Requirement-based verification spanning unit- and component-testing, the MIL, SIL and HIL
  • Data-driven simulation and validation processes spanning Closed-loop and Open-loop methods for
    features and sensors to enable mileage coverage
  • Safety-driven verification and validation to develop
    • test cases against technical safety requirements for software and hardware
    • verification validation via analysis
    • safety and SOTIF scenarios to validate software and sensors against use cases

Leveraging its domain expertise and using high levels of automation, KPIT offers Testing As A Service (TaaS)and incorporates test cases, function-level tests, end-to-end virtual simulations and safety verification and validation for sensor and data cases. It also implements automation and packages — a ready-scenario library for passenger cars and commercial vehicles, automated validation against a ready database of KPIs amongst others — in the data-driven validation area. Despite varied challenges, KPITs automation framework and scenario library accelerates validation coverage. Here’s how.

Computing validation coverage in the virtual simulation

Given how early-testing of AD is unsafe on roads, distances are calculated through simulations. Taking a Confidence Value of 95% and analysing data on possible fatalities contributes to the creation and modelling of different scenarios upon considering the:

  • Mission profile of the vehicle
  • Data for different types of risks – geographical, traffic, weather-related — when there’s a human driver at the wheel
  • Safety goals and hazards
Creating critical, safety and SOTIF scenarios

Closed-loop simulations help validate ADAS and AD features.

Typically SoTIF scenarios are categoriseds as Known and Unknown, Safe and Potentially Hazardous comprising:

  • Area 1: Normal behaviour
  • Area 2: Identified system limitations
  • Area 3: Black swans
  • Area 4: System robustness.

Identifying hazardous faults and failures and ensures the development of a strategy that depends on

  • scenarios for different conditions — low visibility, glaring sun, non-attentiveness of drivers etc.,
  • defining KPIs and
  • the use of Sensor, Traffic, Weather and Environment modelling tools to execute validation end-to-end.
Performing Data-driven Open-loop simulation for sensor validation

Another emergent methodology, the Open-loop simulation or Reprocessing, enables validation of the Camera, Radar and LIDAR. This comprises the:

  • System under test
  • Hardware or HIL system
  • Cloud to store, analyse and manage end-to-end testing

Both Closed-loop and Open-loop simulations for features and sensors are undertaken via a test management suite run on on-premise servers or in the Cloud. KPIT’s comprehensive solution here entails 3 steps:

  • Setup and integration
  • Execution of scenarios
  • Reporting against KPIs to check sensor performance

The Open-loop sensor fusion validation tool is equally key since it validates data converging from different sensors for clearer views of the surroundings. This Open-loop method also uses scenarios different from those validating sensors against signal-level performances for ADAS and AD. Safety validations then, occur at the System, Vehicle and their Integration levels. In this context, the KPIT simulation validation suite spans:

  • Development of systems, libraries with parameters — geographies, highway urban parking, SoTIF etc.
  • Creation of Safety and SoTIF scenarios from real data and strategy for both Closed-loop and Reprocessing methods
  • Their extraction for end-to-end validation

Enabling tech and ecosystems are critical to verification and validation of ADAS and AD features and sensors pre-launch. Testing these features for millions of scenarios is tough. Testing them physically i.e., on the vehicle and the road is expensive and time-consuming. It’s imperative then, that development continue in controlled environments using virtual techniques. Simulation-based virtual validation is effective, efficient and one of the most reliable methodologies available. It must be optimally leveraged.

Copy link