How AnteMotion develops simulation software for autonomous vehicles

AnteMotion, a joint venture of EnginSoft, LHP Engineering Solutions, and V2R - three leading players in research and development for the automotive industry, focuses on developing simulation software for automotive ADAS/AV testing and validation in simulated environments. As soon as they started looking for a location data provider, the team was convinced that HERE meets their needs for HD maps data quality and coverage in Europe and US. We sat down with Matteo Ragni, Senior Mechatronics Systems Engineer, Ph.D who walked us over their development process and the specific challenges simulation software solves for the automotive industry. 

What do you build at AnteMotion and how are your solutions used in the automotive industry?

AnteMotion is building a full pipeline of simulation tools for testing and validation of ADAS / AV applications in virtual environments. The pipeline is composed of three main applications: Maze, Procedural Worlds and Midgard Simulation Framework. Maze is a maps transpiler which generates a rich and detailed representation of the real world. One of the main outputs of Maze is the ASAM OpenDRIVE of an existing area of the world. This file describes both the environment geometries as well as the road network logic. This information is fundamental for several aspects of scenario simulation, such as traffic actors (i.e., vehicles, motorcycles, pedestrians, etc.). The same description is also one of the inputs for Procedural Worlds, the software responsible for procedurally generating stunning virtual 3D environments, full of detail and ready to be used inside Render Engines. The main output of the software is a Universal Scene Description (USD) file. The USD can be imported in the Midgard Simulation Framework, which is built on-top of Unreal Engine 5 providing an astonishingly unique high level photorealistic image quality together with all the functionalities required to virtually test and validate ADAS/AV systems as Software-in-the-loop / Hardware-in-the-loop / Driver-in-the-loop simulations. Beyond the visual superiority, the best thing about the Midgard Simulation Framework is its open framework capability.

How do you use location technology to develop your products? 

AnteMotion's pipeline creates replicas of actual existing road network topologies, mimicking what an ADAS/AV Agents will encounter during operation in the real world. Since the road network is a twin of the real one, real traffic data can be applied in scenario simulation, providing an extremely strong ground to test and validate agents' behavior, even before actual hardware is designed. HERE, with its location technology, enabled this core capability for our pipeline.

We use several HERE APIs, that are fused to obtain a unique and consistent representation of the environment. The 3D geometries of the road are derived from the HERE HD Lane Model APIs, while several assets such as road marks, buildings and traffic signs are derived from HERE SD Maps datasets.  

What challenges did you face while developing your products?

Mapping the real world is complex. Nevertheless, HERE does a very good job; to get a proper geometrical description of the road in OpenDRIVE format, we had to develop a multitude of advanced algorithms to be applied on the input data to obtain a correct geometrical 3D representation of the road network.

We tested several SD and HD Maps providers, and it is evident that HERE has great map coverage in Europe and the United States, something that makes HERE a perfect HD data provider for anyone willing to simulate autonomous vehicles on real road networks.

How does the application help your users? 

One of the main use cases for our users is to replicate complex or tricky traffic situations to virtually test and validate ADAS/AV vehicles’ control and perception systems. As an example, junctions with strange turning angles can rapidly and easily be replicated in a digital environment, and the OpenDRIVE file can be used as a starting point to describe dangerous scenarios (e.g., dangerous pedestrian crossing in hidden corners, in ASAM OpenSCENARIO files) that cannot be tried in real life. With the support of our pipeline, an environment 3D representation, which would require weeks or months of specialized 3D technical artist work, can now be procedurally generated in a matter of hours, boosting effectiveness, and aggressively cutting time to market for our customers.

What else can we expect from AnteMotion in the near future?

We are working on making every aspect of the pipeline even more modular, to allow the customer to personalize each aspect of the environment generation. Support for additional output formats is also one of the next challenges we will tackle, to provide solutions for an even larger set of customers and use cases.