Quebec-based LeddarTech, an automotive software company specializing in AI-based low-level sensor fusion and perception software for ADAS and AD, and TTTech Auto, a technology leader in automotive safety software, have completed a joint demonstrator for a fail-operational architecture using LeddarTech’s LeddarVision software and TTTech Auto’s Safety Co-Pilot solution.
The Safety Co-Pilot solution addresses the challenges of creating a safety architecture suitable for autonomous driving beyond SAE Level 3. It involves a doer channel responsible for the main driving task and an independent checker channel. In the event of an abnormal situation detected by the checker, the doer channel is seamlessly overridden, and a fallback channel brings the vehicle to a safe stop. The collaboration aims to support car OEMs in developing their SAE Level 2-3 safety solutions.
In this collaboration, TTTech Auto leveraged its expertise in safety mechanisms and automotive platforms, complemented by LeddarTech’s advanced fusion and perception capabilities. TTTech Auto provided operational platforms and frameworks to facilitate the development of decomposed architectures for safety applications, while LeddarTech enhanced the partnership with its high-performance fusion and perception stack grounded in low-level fusion.
The joint effort demonstrates an agile approach, using risk assessment and safety-critical event feasibility assessments to present a potential solution. The focus is on showcasing the viability of this approach and its potential for enhancing vehicle safety and cost-efficiency.
LeddarVision is a robust AI-based low-level fusion and perception software stack designed to use raw sensor data from cameras and radars to construct an environmental model. Depending on the sensor set and diversity concept, LeddarVision can be implemented in either the doer or checker channel. TTTech Auto’s Safety Co-Pilot serves as a versatile solution for the checker channel, featuring a set of validation modules that encompass checks related to critical safety requirements. The demonstrator highlights validation modules addressing probabilistic risks associated with physically unfeasible trajectories and potentially dangerous collisions.
The testing of this demonstrator was conducted on data collected from highways in Germany and Israel. The results demonstrate that LeddarTech’s high-quality environmental model and TTTech Auto’s probabilistic checking implement an efficient and reliable checker channel.
Frantz Saintellemy, president, COO and incoming CEO of LeddarTech, said, “Through our partnership with TTTech Auto in the integration of our LeddarVision AI-based software and TTTech Auto’s Safety Co-Pilot, our collective aim has been to showcase a high-level fail-operational architecture suitable for safe driving automation. This collaboration demonstrates our belief that by aligning ourselves with other industry leaders in win-win partnerships, we accelerate technology deployment and adoption.”
