How Software Tools Can Advance Autonomous Vehicle Development

Here’s the deal breaker regarding autonomous vehicles (AV): They have to drive better than human drivers.

Taking the human driver out of the loop complicates AV validation, making the endeavor difficult and time-consuming. And yet there’s basically no other way. Validating AV technology is nearly impossible through physical road testing alone. For instance, Akio Toyoda, president of Toyota Motors, has been quoted as saying that fully autonomous vehicles would have to be driven 8.8-billion miles to ensure their safety.

Software, which is at the core of AV technology, is also at the core of designing, training, testing, and certifying that technology. This software runs the gamut, from design and analysis, to software- and hardware-in-the-loop testing, to simulation, to virtual reality, and much more. Keep in mind: No single software vendor provides it all—not even ANSYS Inc. (ansys.com), the focus of this selective round-up of software for AV development.

Start with safety-critical analysis
 

AV development at ANSYS starts with functional safety analysis in accordance with applicable safety standards (e.g., ISO 26262 and ISO/PAS 21448), explains Sandeep Sovani, global director of automotive & ground transportation for ANSYS. Specifically, ANSYS medini analyze software automates the analysis and verification of functional safety for electronic control systems; helps implement step-by-step modeling, analysis, and verification processes; helps ensure the implementation of safety mechanisms; and maintains traceability of requirements and design modifications across the AV development process. ANSYS medini graphically models functional architecture, along with functional dependencies and malfunctions. Other main functions include table-based management of driving situations and hazardous events; quality analysis for product design and related processes; driving situation, hazard and risk, fault tree, failure mode and effects, and probabilistic analyses; customizable work product/documentation generation; and integration with third-party requirements management software.

Generating software
 

ANSYS SCADE (Safety Critical Application Development Environment) is for model-based embedded software development and embedded controls—the software that operates AVs and subsystems, such as head-up displays (HUD). The SCADE suite covers requirements management, model-based design, simulation, verification, qualifiable/certified code generation, software error detection and testing, and interoperability with other software development systems. SCADE features legacy code imports into designs; semantic comparisons of models, operators, and state machines; parallel software architecture and design development; bi-directional synchronization between architecture models and design models; and analysis of worst-case execution time. The SCADE Suite Automotive package provides AUTOSAR R 4.2.2 support, calibration and fixed-point support (ASAM MCD-2 DC), certified code generation by TÜV SÜD at ISO 26262 TCL3 for ASIL D Software Development.

SCADE does not generate neural net/artificial intelligence software. For that, SCADE can build the “wrappers” around neural networks to create safety-compliant processes.

Validating sensors
 

Sensors are the “eyes and ears” of AVs. The sensors for viewing the external world, include cameras, radar, ultrasonic, and lidar. These sensors are based on different technologies, requiring different software tools for development and validation.

ANSUS SPEOS optical simulation software uses high-fidelity physics simulations to model and analyze the luminous flux of ultraviolet to near-infrared optical systems (in AVs, mostly optical cameras and lidars). SPEOS determines the visual aspect, reflection, visibility, and information legibility of objects and interior views of a vehicle from a person’s perspective.

For radar sensors, ANSYS HFSS uses the physics of electromagnetics (EM) to simulate such characteristics as antenna radiation patterns, signal reflections during propagation and reception, signal strengths, and obstructions from objects in and outside the passenger cabin. HFSS also lets engineers extract parasitic parameters, visualize 3D EM fields (near- and far-field), and generate full-wave SPICE models linked to circuit simulations.

For the ultrasonic sensors in short-distance sensing, ANSYS has its traditional ANSYS Mechanical software, which includes acoustic analysis.

These software tools are also used for integrating the sensors in the vehicle itself. ANSYS Mechanical, for example, is used for analyzing thermal and structural factors, such as whether the sensors can withstand the heat of the sun for hours on end and the harsh bumps and vibrations of normal vehicular operation.

Computers abound
 

To ensure the reliability of semiconductors, chips, and electronics in AVs, plus electric motors, engineers can turn to a slew of ANSYS software tools. Here’s an incomplete list:

• Electronics Desktop for designing and simulating electrical, electronic and EM components, devices, and systems while accessing thermal, fluid, and mechanical software for multiphysics analyses
• Icepak, which uses Fluent computational fluid dynamics, for thermal and fluid flow analyses of integrated circuits (ICs), printed circuit boards (PCBs), and electronic assemblies
• Q3D Extractor for signal-integrity analysis of EM phenomena and performance analysis of interconnects, ICs, connectors, PCBs, bus bars, and cables
• Maxwell for EM field simulation
• Path FX for evaluating delay and variance in system on a chip (SoC) components
• PathFinder to analyze SoC designs subject to electrostatic discharge (ESD)
• Pharos for identifying EM and substrate crosstalk net/block
• RedHawk for analyzing high-performance and power-efficient SoCs in harsh environments (e.g., thermal, EM, and ESD).

The all-important interface
 

Until cars are fully autonomous, some connection needs to exist between driver and vehicle, says Sovani. Specifically, he’s talking about human-machine interfaces (HMI) at the lower levels of autonomy before humans can simply give AVs verbal instructions. SCADE Display, SPEOS, and ANSYS VRXPERIENCE can all be used for developing and testing HMI, such as evaluating HUD presentations of vehicle operational data and whether displays are bright enough.

There’s more
 

“Cyber security is a special safety case,” explains Sovani. “These are not passive infractions, but active hacking attempts.” ANSYS medina Cyber Security Package can analyze possible attack trees, and explore system vulnerabilities.

ANSYS relies on its software partners for application lifecycle management (ALM). However, ANSYS EKM (Engineering Knowledge Manager) provides simulation process data management, including curating simulation data, performing metadata analysis, and connecting to ALM and product lifecycle management (PLM) packages. EKM provides real-time traceability and audit trails based on applicable regulatory compliance; extensive reporting; advanced keyword, meta-data, property, and report-based search, with filtering based on object type, subprojects, and so on; visualization for all data types; and the ability to create, edit, and publish simulation workflows and lifecycle definitions.

Validating through simulation
 

Validating the AV software, explains Sovani, is “typically done by putting software into a car, letting the software drive the car on public streets—with a driver behind the wheel—and seeing what decisions the software is making.” This level of simulation is easier, faster, safer, and far more complete than on-the-road, human-driven testing.

Or one can simulate the entire car. By virtualizing systems, subassemblies, hardware, and software in VRXPERIENCE, engineers can perform closed-loop simulations of driving based on the physics-based simulations of the sensors “on” the vehicle. For example, a simulation of a camera sensor can be connected to the camera electronic control unit (ECU) for a hardware-in-the-loop simulation to test and validate the ability of the perception software on the ECU chip to correctly interpret images generated by the simulated camera sensor.

Continues Sovani, “The software thinks it’s driving the real car in the real world, but it’s actually driving the virtual car in the virtual world.” Which is a safe approach to validating hardware and software. Sovani explains, “Who cares if a virtual car hits a virtual pedestrian in a virtual world? But obviously we care in the real world.” So simulation provides a real-world benefit.

Just as important, simulation expedites the testing process. Consider all the common and conceivable variations when driving from Point A to Point B: road surface and path, weather, traffic (pedestrian and other vehicles), sun location (rising or setting), animal crossings, plastic bags skittering across the road, to name a few. “All these variations can be run quickly and automatically to make sure that the software is doing the right things,” says Sovani.

Not just the right things, but the safe things.