The newly launched LVF-H is a comprehensive fusion and perception software stack supporting premium front-view L2/L2+ ADAS highway assistance and 5-star NCAP 2025/GSR 2022 safety applications. LVF-H implements a premium stack handling sensors’ interface, offline and online calibration and diagnostics, sensor synchronization, sensor fusion, object detection and classification, extended to include unclassified objects and events (e.g., cut-in), continuous tracking and stabilization, free space detection, road model, comprehensive traffic signs detection, highway traffic light detection, vehicle odometry interface, ego-motion localization and global localization with HD Map input. Stack further provides ability to fuse external sensors such as HD Map and V2X into a single and unified environmental model. Stack extends safety features support with detected objects trajectory prediction, perception decomposition, perception support to ODD analysis, sensor coverage and health monitoring. The complete stack extends perception, positioning and prediction features providing a comprehensive single and unified environmental model and API to premium L2/L2+ and highway assist ADAS applications.
Features and Benefits
Based on LeddarVision™ low-level fusion and perception technology, it enables the premium L2/L2+ and highway assist ADAS offering at lower sensor and hardware cost. Low-level fusion optimally combines sensor modalities, pushing performance envelope far beyond legacy solutions and extending supported object detection fusion range over 200 m. LVF-H extends the front camera-based configuration to a 1V5R configuration, having a single wide FoV (120°) 3 Mpx class front camera, one medium-range front radar and four short-range front and back corner radars. Front camera and medium-range radar extends the fusion range over 200 m and ACC support up to 160 km/h. The back short-range corner radars extend the support of GSR 2022 and 5-star NCAP 2025 to overtaking/reverse/dooring scenarios and the highway assist features support to lane change collision warning. LVF-H software stack is targeting low-cost ECU based on TI’s TDA4L plus a single Hailo-8 AI processor offering together over 30 TOPS of deep-learning acceleration to support the premium perception.
LVF-H superior object detection performance further extends to occluded VRUs and vehicles, providing early warning in NCAP tested scenarios (e.g., occluded cyclist tests). Superior object separation and longitudinal position measurement accuracy in highway scenarios leads to higher performing ACC implementation. Inherent sensor redundancy provides a more reliable operation in presence of degraded (e.g., dirty lens), failing (e.g., failing camera) or conflicting sensors (e.g., false alarms from radars in presence of guard rails’ multipath) and adverse scenarios and environments (e.g., dust, blinding light, etc.).
LVF-H is part of a family of front-view products targeting entry- to premium-level ADAS. A comprehensive roadmap provides growing features support for different market segments. LVF-E companion product targets entry-level front-view ADAS, with lowest-cost 1V2R sensor configuration, having only a single wide FoV (120°) 1.2 Mpx class front camera and two short-range front corner radars to support GSR and five-star NCAP 2025 with comprehensive perception and positioning features.