Semantic Grid Map based LiDAR Localization in Highly Dynamic Urban Scenarios 12 th IROS20 Workshop on Planning, Perception and Navigation for Intelligent Vehicles Oct. 2020 Chenxi Yang, Lei He, Hanyang Zhuang, ▪ Chunxiang Wang, Ming Yang * ▪ * mingyang@sjtu.edu.cn ▪ ▪ This work is supported by the National Natural Science Foundation of China (U1764264/61873165)
Contents Intr troducti duction on 1 Related ated wo work k 2 Se Semantic tic grid id map 3 4 Local alization ization Experi riment ent 5
Introduction Loca caliza ization ion Planni nning ng Perce ceptio ption Key tech. Ke h. of autonom nomous ous driving iving Navigatio gation Contr trol ol
Introduction Localiza Lo alization tion in AD Dynam namic ic GNSS SS Map-bas ased ed pose estim imation ation interferen rferences ces Signal denial Illuminance changes Robust to illuminance Multipath effect Low reliability High reliability
Contents Intr troducti duction on 1 Related ated wo work k 2 Se Semantic tic grid id map 3 4 Local alization ization Experi riment ent 5
Related work Environ ironme menta ntal l ma mapp pping Map Map fo form GNSS-based Point sed nt cloud oud map ✓ global consistency ✓ accuracy • • signal denial data size • real-time performance 2D 2D gr grid d map ✓ data size & speed • information lost SLAM-base Feature sed re map ✓ local consistency ✓ accuracy & speed • • cumulative error sensitive to the environment
Represent esentati ative ve methods ods Agency cy Challe lenges ges Non-real time IMLS-SLAM Point MINES (ICRA2018) Dynamic interference Probabilistic Maps Stanford (ICRA2010) Grid Dynamic interference Robust Localization Baidu (ICRA2017) Non- semantics LOAM CMU (ICRA2014) Feature Dynamic interference SuMa Bonn (ICRA2018 ） Segmatch LiDAR-based ETH (RSS2018) Non-real time Descriptor L3-Net localization Baidu Dynamic interference (CVPR2019) Lane Semantic KIT Feature missing (IROS2018) Pole-like feature Freie Robustness (IROS2016 ） Semantics Point cloud Semantic ICP Non-real time UMich (BMVC2018) segmentation Non-semantics: dynamic interference Semantics: difficult to find a balance between real-time and robustness Multiple semantic features
Contents Intr troducti duction on 1 Related ated wo work k 2 Se Semantic tic grid id map 3 4 Local alization ization Experi riment ent 5
Semantic grid map ▪ Featur ture e selecti ection on ▪ Abundant in urban scenarios ▪ Strongly imply static ▪ Extractable from scan-level sparse point cloud ▪ Sufficient pose constraints from multiple layers ▪ Seman antic tic grid d map ▪ To speed up the calculation ▪ Semantic category with a trust rate
Contents Intr troducti duction on 1 Related ated wo work k 2 Se Semantic tic grid id map 3 4 Local alization ization Experi riment ent 5
Localization ▪ On On-li line ne pose se initiali ializa zation tion ▪ Large range search ▪ Limited to the first several frames ▪ Relatively low real-time requirements -> to keep as much map detail as possible, the SGM is in 3D formed by cubes ▪ Real-tim time e traje jecto ctory ry tracki cking ng ▪ Can inherit an accurate initial position from the previous frame ▪ Every frame ▪ Strict real-time requirements (typically 100ms) -> to ensure the calculation speed, the SGM is in 2D formed by squares
Localization ▪ On On-li line ne pose se initiali ializa zation tion ▪ Notation Map Cubes Scan Cubes GMM Semantic Category
Real-time trajectory tracking ▪ Real-tim time traje jecto ctory ry tracki cking ▪ Notation ▪ Residual error
Contents Intr troducti duction on 1 Related ated wo work k 2 Se Semantic tic grid id map 3 4 Local alization ization Experi riment ent 5
Experiment ▪ Proces cesso sor ▪ Intel i7-7567U @3.5GHz with 16GB memory ▪ Express press road ▪ 5.2km long
Experiment ▪ On On-li line ne pose se initiali ializa zation tion ▪ (0.2m) 3 cube ▪ horizontal offset uniform distribution in 50m circle ▪ up to 90 degree offset ▪ a s speci cial al case Conjuncti nction on CPD Result lt 2 nd d iterat ation on Initial ial position ion Ours
Experiment ▪ Real-tim time traje jecto ctory ry tracki cking ▪ (0.1m) 2 square
Experiment ▪ Proces cesso sor ▪ Intel i7-7567U @3.5GHz with 16GB memory ▪ Facto ctory ▪ 1.5km long
Thank you for your attention!
Recommend
More recommend
