Towards automatic driving Collective Perception Dr. Teodor Buburuzan, Hendrik-Jörn Günther 10 th March 2016
V2X Roadmap Applications Take-over of the driving functions 100 80 60 40 20 0 Status Data Emergency Vehicle W Dangerous Situation W Stationary Vehicle W Traffic-Jam W Pre-/Post-Crash W Hazardous Location W Adverse Weather W Roadworks W 1.0 Sensor Data GLOSA 1.0 In-Vehicle Information Roadworks W 2.0 Connected ACC Overtaking W Intersection Collision W... Intention Data GLOSA 2.0 Roadworks Assistance Lane-Merge Assistance Area Reservation Cooperative ACC VRU Warning Platooning Automation Level Coordination Data Cooperative Merging Overtaking Assistance Intersection Assistance Dynamic Platooning VRU Assistance Fully Automated Driving Optimal Traffic Flow 100% installation of new vehicle sales 100% installation of new vehicle plattforms 10 year ramp-up to 100% installation of new vehicles Installation on select new vehicle type of luxury and upper middle class vehicles Phase 1 Awareness Driving Phase 2 Sensing Driving Phase 3 Cooperative Driving Phase 4 Synchronized Cooperative Driving Phase 5 Accident-free Driving Dissemination Cooperation 10.03.2016 WG Roadmap 2
V2X Roadmaps guiding principles inside C2C-CC Focus on information exchange Cooperate on providing information Provide clearly defined/specified information Use commonly agreed air-interfaces Focus on localized dissemination patterns Use modular information which builds on top of each other Compete in capitalizing on the information Each traffic participant can (freely) use the received information Automatic driving functions can be enhanced (range, success rate) 10.03.2016 WG Roadmap 3
Levels of Cooperative Systems Which type of information is exchanged between vehicles? Planning Coordinated Driving Maneuvers Plans for us Intention Information My intention and/or proposals for you 4 3 Perception Environmental Information My environment Status Information My status 2 1
Motivation Local Perception Sensors Red ego-vehicle knows about objects detected by its local perception sensors 5
Motivation Local Perception Sensors CAM CAM CAM CAM V2X Communication - Red ego-vehicle additionally knows about V2X vehicles within the communication range - Vehicles broadcast Cooperative Awareness Messages (CAMs) CAM 6
Motivation Local Perception Sensors EPM V2X Communication EPM EPM Collective Perception - V2X vehicles with local perception sensors broadcast their locally perceived objects - Vehicles broadcast Environmental Perception Messages (EPMs) 7
Methodology Scenario: 60% penetration rate Every V2X enabled vehicle is equipped with front- and rear- facing radar sensors CAM dissemination according to ETSI standard for every V2X vehicle Every vehicle within the local perception range of a V2X vehicle is broadcasted by an EPM [Environmental Perception Message] at a constant rate of 1 Hz Simulations: 10 scenarios @ 7 different penetration rates 70 simulations with 2218 vehicles per variation of penetration rate Non-V2X-enabled V2X-enabled 8
Methodology Awareness Ratio 94 Description: Communication range for ego-vehicle is displayed (300 m reference range) In the vicinity of the ego-vehicle (ID 43) are several V2X vehicles as well as non-v2x vehicles (IDs 0, 17, 88, 100) 65 7 Concept: As a measure for the effectiveness of collective perception, the metric awareness ratio kk rrrrrr is introduced: 82 80 84 81 50 41 59 36 46 17 0 107 9 88 43 62 70 98 79 Possibility of receiving vehicles outside of the communication range: kk rrrrrr 100 Non-V2X-enabled V2X-enabled Ego 9
Methodology Awareness Ratio 94 Description: Communication range for ego-vehicle is displayed (300 m reference range) In the vicinity of the ego-vehicle (ID 43) are several V2X vehicles as well as non-v2x vehicles (IDs 0, 17, 88, 100) Ego-vehicle perceives all V2X objects within communication range kk CCCCCC rrrrrr = 9 = 0.69 13 82 80 84 81 50 41 59 36 46 17 0 107 9 88 43 62 70 98 79 65 7 100 Non-V2X-enabled V2X-enabled Ego perceived by CAM 10
Methodology Awareness Ratio 94 Description: Communication range for ego-vehicle is displayed (300 m reference range) In the vicinity of the ego-vehicle (ID 43) are several V2X vehicles as well as non-v2x vehicles (IDs 0, 17, 88, 100) Ego-vehicle perceives all V2X objects within communication range kk CCCCCC rrrrrr = 9 = 0.69 13 Locally perceived objects by radar: kk RRRRRRRRRR rrrrrr = 4 = 0.31 13 82 80 84 81 50 41 59 36 46 17 0 107 9 88 43 62 70 98 79 65 7 100 Non-V2X-enabled V2X-enabled Ego perceived by CAM perceived by radar 11
Methodology Awareness Ratio 94 Description: Communication range for ego-vehicle is displayed (300 m reference range) In the vicinity of the ego-vehicle (ID 43) are several V2X vehicles as well as non-v2x vehicles (IDs 0, 17, 88, 100) Ego-vehicle perceives all V2X-objects within communication range kk CCCCCC rrrrrr = 9 = 0.69 13 Locally perceived objects by radar: kk RRRRRRRRRR rrrrrr = 4 = 0.31 13 Knowledgedge due to exchange of locally perceived objects (EPM): kk EEEEEE rrrrrr = 11 = 0.85 13 82 80 84 81 50 41 59 36 Non-V2X-enabled V2X-enabled Ego perceived by CAM perceived by radar perceived by EPM 46 17 0 107 9 100 88 43 62 70 98 79 65 7 12
Methodology Awareness Ratio 94 Description: Overall awareness ratio : kk tttttt rrrrrr = uuuuuuuuuuuu IIIIII vvvvvvvvvvvvvvv wwwwwwwwwww cccccccccccccccccccccccccc rrrrrrrrrr 65 7 = 0,9,17,46,62,65,70,79,81,88,98,107 0,9,17,46,62,65,70,79,81,88,98,100,107 = 12 13 = 0.92 82 80 84 81 50 41 59 36 46 17 0 107 9 88 43 62 70 98 79 Non-V2X-enabled V2X-enabled Ego perceived by CAM perceived by radar perceived by EPM 100 13
Methodology Awareness Ratio 94 Description: Overall awareness ratio : kk tttttt rrrrrr = uuuuuuuuuuuu IIIIII vvvvvvvvvvvvvvv wwwwwwwwwww cccccccccccccccccccccccccc rrrrrrrrrr 65 7 = 0,7,9,17,36,46,62,65,70,79,81,84,88,94,98,107 0,9,17,46,62,65,70,79,81,88,98,100,107 = 16 13 = 1.23 It is possible to gain information about vehicles outside of the communication range Advantageous in situations of limited and obstructed communication range Redundant information 82 80 84 81 50 41 59 36 46 17 0 107 9 88 43 62 70 98 79 Non-V2X-enabled V2X-enabled Ego perceived by CAM perceived by radar perceived by EPM 100 14
Findings Awareness ratio over time 1 Awareness ratio (300 m range) 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 Day-2 scenario (Radar & CAM & EPM) Day-1 scenario (Radar & CAM) Radar - µ : 0.178, σσ : 0.0235 CAM - µ : 0.61, σσ : 0.0244 Radar & CAM - µ : 0.659, σσ : 0.0252 EPM - µ : 0.792, σσ : 0.0827 Radar & CAM & EPM - µ : 0.96, σσ: 0.1 100 110 120 130 140 150 160 170 180 190 200 simulation time 15
Findings Awareness ratio from Radar 1 0.9 0.8 Independent of the penetration rate, the relative knowledge for objects perceived by radar is almost constant Outliers are very specific to traffic scenarios (i.e. vehicles located around a junction) awareness ratio for 300 m range 0.7 0.6 0.5 0.4 0.3 0.2 Radar ( µ ) 0.1 0 5 10 20 40 60 80 100 penetration rate in % 16
Findings Awareness ratio from CAMs 1 0.9 0.8 Awareness ratio for objects perceived by CAM is equal to the penetration rate Depending on the traffic scenario, higher local penetration rates can be observed awareness ratio for 300 m range 0.7 0.6 0.5 0.4 0.3 0.2 Radar ( µ ) CAM ( µ ) 0.1 0 5 10 20 40 60 80 100 17
Findings Awareness ratio from EPMs 1 0.9 0.8 When rebroadcasting objects that could be perceived by both, V2X and local sensors, the perception range can be extended beyond the communication range awareness ratio for 300 m range 0.7 0.6 0.5 0.4 0.3 Part of EPM Not part of EPM Even at a penetration rate of 100%, not all vehicles can be perceived by local sensors (only) 0.2 0.1 Radar ( µ ) CAM ( µ ) EPM ( µ ) 0 5 10 20 40 60 80 100 penetration rate in % 18
Findings Combined awareness ratio 1 0.9 0.8 Combination of all three sources (local sensors, CAM and EPM) results in leverage for all penetration rates Especially for low penetration rates, collective perception unveils its largest potential awareness ratio for 300 m range 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Radar ( µ ) CAM ( µ ) EPM ( µ ) Radar & CAM & EPM ( µ ) 0 5 10 20 40 60 80 100 19
awareness ratio for 300 m range Findings Relative knowledge standard deviations 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 CAM: mediocre variance Radar: potential relative awaress specific to scenario EPM: Very specific to scenario: Even at high penetration rates, 100% awareness based on radar cannot be reached as there may be situations with no vehicles in local perception range 0.2 0.1 Radar ( µ ± σ ) CAM ( µ ± σ ) EPM ( µ ± σ ) Radar & CAM & EPM ( µ ± σ ) 0 5 10 20 40 60 80 100 penetration rate in % 20
Conclusion Enables vehicles to get a more complete picture of their environment At low penetration rates offers considerable enhanced awareness ratios Additional views of individual objects might prove helpful with functional safety issues. Collective Perception is a crucial step towards fully automatic-driving! 21
V2X Roadmap Technology Phase 1 Phase 2 Phase 3 Phase 4 Release / time Simultaneous multi-channel 60 GHz Comm. Simultaneous dual-channel G5D-SC5 G5B-SC4 G5B-SC3 G5A-SC2 G5A-SC1 G5A-CCH Day1 CC single-channel switched-mode Segmentation/Reassembly Data Streaming Advanced FWD GN-Groupcast GN6 GN/BTP + QoS GN-Unicast multi-channel CC dual-channel dual-channel CC GN-GBC GN-SHB Symmetric Crypto. Misbehaviour detection Crypto-Agility PC-change rules Day-1 PKI Automatic-driving Messages I2V Coop. Messages Intention Msg. Platoon Control Msg. Platoon Management Msg. Electronic Horizon Msg. IVI-Platoon Extensions Collective Perception Msg. Collective Positioning Msg. CAMv2 SAM Parking Mng. Msg. Digital Inspection Msg. SPAT MAP IVI DENM CAM Automation Level Dissemination Cooperation domain 10.03.2016 WG Roadmap 22
Thank You! Questions? Dr.-Ing. Teodor Buburuzan teodor.buburuzan@volkswagen.de Hendrik-Jörn Günther hendrik-joern.guenther@volkswagen.de
Sensor Properties V2X communication range (reference range) 60 60 80 m Local perception sensor (local perception range) 80 m