Grant: Discovery Grants Program - Individual
University of Calgary
$32,000.00
Factor graph based high-definition mapping system for autonomous driving
EF 2023-24
Résumé
FACTOR GRAPH BASED HIGH-DEFINITION MAPPING SYSTEM FOR AUTONOMOUS DRIVING With the increasing market demands from various applications such as autonomous driving and smart city building, high-definition (HD) maps attract more and more attention from both industry and academia. However, there are still gaps need to be filled for the accuracy, reliability, efficiency and scalability of the currently existing HD mapping algorithms and pipelines. Over five years, this research program will focus on the HD mapping system development for autonomous driving, which includes map pose estimation module, map content generation module, map updating and map validation module. For map pose estimation, the information from various sensors including global navigation satellite system (GNSS), inertial measurement unit (IMU), light detection and ranging (LiDAR), camera, radio detection and ranging (RADAR) and odometer will be integrated to ensure high accuracy and reliability. First of all, evaluation and modelling will be carried out for both the navigation sensors (GNSS, IMU and odometer) and perception sensors (LiDAR, RADAR and camera). Afterwards, new sensor calibration process and algorithm will be developed based on a single dataset, which is called all-in-one calibration by computing extrinsic parameters efficiently. As an advanced optimization method, the factor graph is proposed for tightly coupled sensor fusion in map pose estimation, which could ensure the high accuracy and reliability of the map pose. On top of the map pose estimation, the map content generation will be implemented based on deep learning. Models will be trained end-to-end with transformer to generate rich map features. Meanwhile, the HD mapping algorithms on snowy days will be developed to extend the operation design domain (ODD) of autonomous driving with real data collected in Canada. Map updating algorithm based on low-cost sensors will be developed to maintain a large-scale HD map. Meanwhile, the pose simultaneous localization and mapping (SLAM) algorithm will be developed for the point cloud registration and image registration between the base map and new collected map to ensure consistency. For map validation, both the offline and online validation algorithms will be developed. The offline algorithm will compare the new map and base map in pose domain and map feature domain. The validation will also be carried out based on the online localization module in real-time, which could provide efficient feedbacks regarding the accuracy and effectiveness of map features in the HD map. The proposed research program focused on HD mapping will bring abundant benefits to Canadian society, including advancement of the highly autonomous driving development, smarter and safer traffic system, lower fuel use and carbon emission, better accessibility for seniors and people with disabilities, new business opportunities such as goods delivery.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Alberta, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
RGPIN-2023-03765
Name nom
Yang, Hongzhou
Department département
Department of Geomatics Engineering
Organization id
11
Institution établissement
University of Calgary
Province en
Alberta
Province fr
Alberta
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2023
Award amount
32000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1509
Committee name en
Civil, Industrial and Systems Engineering
Committee name fr
Génie civil, industriel et des systèmes
Area of application code
701
Area of application group en
Transportation systems and services
Area of application group fr
Services et systèmes de transport
Area of application en
Ground (road and rail)
Area of application fr
Terrestre (routier et ferroviaire)
Research subject code
9999
Research subject group en
Not available
Research subject group fr
Not available
Research subject en
Non disponible
Research subject fr
Non disponible
Application title
Factor graph based high-definition mapping system for autonomous driving
Keyword
All-in-one calibration; Autonomous driving; Deep learning; Factor graph; High-definition map; Map content generation; Mapping on snowy days; Map update with low-cost sensors; Map validation; Sensor fusion
Application summary
FACTOR GRAPH BASED HIGH-DEFINITION MAPPING SYSTEM FOR AUTONOMOUS DRIVING With the increasing market demands from various applications such as autonomous driving and smart city building, high-definition (HD) maps attract more and more attention from both industry and academia. However, there are still gaps need to be filled for the accuracy, reliability, efficiency and scalability of the currently existing HD mapping algorithms and pipelines. Over five years, this research program will focus on the HD mapping system development for autonomous driving, which includes map pose estimation module, map content generation module, map updating and map validation module. For map pose estimation, the information from various sensors including global navigation satellite system (GNSS), inertial measurement unit (IMU), light detection and ranging (LiDAR), camera, radio detection and ranging (RADAR) and odometer will be integrated to ensure high accuracy and reliability. First of all, evaluation and modelling will be carried out for both the navigation sensors (GNSS, IMU and odometer) and perception sensors (LiDAR, RADAR and camera). Afterwards, new sensor calibration process and algorithm will be developed based on a single dataset, which is called all-in-one calibration by computing extrinsic parameters efficiently. As an advanced optimization method, the factor graph is proposed for tightly coupled sensor fusion in map pose estimation, which could ensure the high accuracy and reliability of the map pose. On top of the map pose estimation, the map content generation will be implemented based on deep learning. Models will be trained end-to-end with transformer to generate rich map features. Meanwhile, the HD mapping algorithms on snowy days will be developed to extend the operation design domain (ODD) of autonomous driving with real data collected in Canada. Map updating algorithm based on low-cost sensors will be developed to maintain a large-scale HD map. Meanwhile, the pose simultaneous localization and mapping (SLAM) algorithm will be developed for the point cloud registration and image registration between the base map and new collected map to ensure consistency. For map validation, both the offline and online validation algorithms will be developed. The offline algorithm will compare the new map and base map in pose domain and map feature domain. The validation will also be carried out based on the online localization module in real-time, which could provide efficient feedbacks regarding the accuracy and effectiveness of map features in the HD map. The proposed research program focused on HD mapping will bring abundant benefits to Canadian society, including advancement of the highly autonomous driving development, smarter and safer traffic system, lower fuel use and carbon emission, better accessibility for seniors and people with disabilities, new business opportunities such as goods delivery.