Mobileye’s Executive VP of Autonomous Driving, Johann Jungwirth, outlined the company’s methodical approach to sensor integration in autonomous vehicles. Plan demonstrates Mobileye’s commitment to what they’re calling a “vision-first” foundation while maintaining stringent safety standards through strategic redundancy.
“We’re a vision-first company, and computer vision is our core foundation,” Jungwirth explained during a recent interview with S&P Global Mobility. Philosophy underpins Mobileye’s entire autonomous driving platform.
The company’s sensor implementation follows a carefully calibrated progression. Self-driving system powered by the EyeQ6 SoC employs full 360-degree coverage across three sensor types: cameras, self-developed 4D imaging radar, and LiDAR. Configuration ensures triple redundancy, particularly at the vehicle’s front, the most critical perception zone.
In the second phase, Mobileye will strategically reduce LiDAR usage to front-only implementation, eliminating side and rear units. Reasoning behind this adjustment is practical: imaging radar doesn’t require cleaning systems like LiDAR does and lacks the moving parts that can compromise durability in mechanically-structured LiDAR units.
“Safety is our top priority, guiding every architectural decision,” Jungwirth emphasized.
Current Mobileye stack features 13 cameras, comparable to Waymo’s 6-generation system, alongside five imaging radars. Approach differs slightly from competitors, where Waymo positions two smaller radars at front corners, Mobileye deploys a single large 1536-virtual-channel imaging radar up front, complemented by smaller radars around the vehicle.
What makes Mobileye’s three-tier sensor strategy particularly notable is its data-driven approach. The company isn’t making speculative technology bets but instead following empirical evidence regarding sensor efficacy.
Mobileye’s software-defined 4D imaging radar represents a significant technological advancement. Can rapidly switch between ultra-short, short, mid, and long-range modes in real-time, producing point clouds with resolution approaching LiDAR quality.
Jungwirth expressed enthusiasm about future developments: “We’re excited for advances in SoC efficiency and antenna design to unlock more resolution, range, dynamic range, and performance.”
Beyond hardware diversity, Mobileye implements additional safety frameworks including their Primary-Guardian-Fallback methodology, Responsibility-Sensitive Safety driving policy, and REM mapping technology.
Third step in Mobileye’s three-tier sensor strategy represents perhaps its most forward-looking position: the theoretical possibility that even front LiDAR might eventually become unnecessary. However, Jungwirth carefully qualified this projection, noting it would be “strictly determined by future data.”
Related Post
Mobileye Ends Lidar Development, Laying Off about 100 Employees
Beyond Lidar: How Tesla Vision-only E2E and MobilEye Radar Are Reshaping Autonomous Driving
Mobileye CEO Questions Viability and Safety of Tesla’s End-to-End FSD Approach