GNSS in Autonomous Vehicles MM Vision MM Technology Innovation Automated Driving Technologies (ADT) Evaldo Bruci
Context & motivation Within the robotic paradigm Magneti Marelli chose Think & Decision as the key item for study, development, prototyping, product planning Our framework platform is called: Cognitive Fusion Framework (CFF) 2
Automated Driving: Architecture 2015 2016 2017 2018 Distributed Mixed Centralized Smart Smart Smart VDe 1.0 Smart VDe 2.0 Smart Actuation Actuation 3
From perception to movement Cognitive Fusion Framework Perimeter Actuation Layer Vehicle Control Decision Layer Planning & Decision (Action & Path) Data Fusion Layer Data Fusion Fusion Layer Perception Layer Fusion Lidar Camera Radar Ultrasounds Environmental Awareness Behavioral Prediction Maps & Positioning GPS IMU Visual Odometry Maps [Ref. JDL Model] 4
Fusion to create an environmental model & Data Fusion Environment Awareness Behavioral prediction Source: http://spectrum.ieee.org/cars-that-think/transportation/selfdriving/google-autonomous-cars-are-smarter-than-ever 5
Decide if and how to move Action Planning Keywords Risk Assessment Deep Neural Networks Safety Fuel Efficiency Comfort Path Planning Keywords Trajectory estimate 6
Localize with high precision Precise GNSS Map Manager & Precise Positioning Precise Inertial sensors HD Maps Lane detection/visual map matching 7
Magneti Marelli PoC demonstrated in 2016 Highway use case Automatic Valet Parking use case 8
Magneti Marelli PoC under development Urban use case Long distance Highway use case 9
Precise Positioning - Precise map: Function It must works in fusion with the whole system. But in failure mode it must be able to work alone. Positioning+map must: Guarantee navigation among the parking spaces in automatic valet parking Guarantee lane keeping in motorway, highway and city boulevards 10
Precise Positioning Precise map: Target accuracy Target accuracy: Positioning accuracy : 20 cm Cartography accuracy: 20 cm max error on node coordinates Strategy for achievemnt: Positioning accuracy : Accurate sensors Well designed sensor fusion GNSS Ultra tight integration GNSS PPP RTK Visual Map Matching Cartography: Work on requirements with suppliers 11
GNSS Phase detection Strategy Pro s Con s PPP Function suitable for automotive Subscription price not yet at automotive level Receiver price not yet at automotive level RTK Require additional communication channel Low-zero price available, but not world wide spread Receiver price not yet at automotive level 12
GNSS Cost consideration COST IS THE BLOCKING FACTOR Reasonable autonomous car price 30-50 Keuro The cost of a GNSS with phase detection capability is still between 1 and 10 KEuro Automotive cost is the key issue for GNSS success in autonomous driving Receivers Services 13
GNSS Cost consideration The strategy is cooperation with suppliers: GNSS engine suppliers PPP RTK corrections providers Points to be exploited : Accurate definition of the products specification s Advantage of automotive level mass production 14
Precise Positioning Ultra tight integration Highly promising in terms of GNSS reliabilty/integrity Accuracy increase is espected No added cost in the final product 15
Precise Positioning Visual map matching An autonomous vehicle must have Cameras/Laser scanners No added cost in the final product 16
Thank you