Embark SPAC Presentation Deck

Vision Map Fusion & Scene Understanding Enable Robustness Moving Beyond Brittle, Map-Centric Approaches Vision Map Fusion Using proprietary techniques, Vision Map Fusion uses a non-linear optimization approach to fuse local vehicle poses with global map geometry in real-time, enabling on-the-fly map corrections Making map corrections in real-time allows for dramatically improved accuracy and robustness of near-field lane geometry estimation, while also unifying the corrected near-field topology with the far- field map to create a static scene used in downstream operations Our approach is designed to be less map reliant and more capable of handling the constantly changing road environment By correcting map errors on the fly, Vision Map Fusion can facilitate safer, more scalable, and more efficient operations compared to alternative map-centric approaches W NORTHERN GENESIS II Grady Williams ● PhD Robotics from Georgia Tech Creator of MPPI (Model Predictive Path Integral) algorithm Widely cited in vehicle control, robotics and other research 12

View entire presentation