SLAM · Localization · Mapping · Multi-Sensor Context Fusion · Reinforcement Learning
Chengcheng Guo
Senior Engineer at NVIDIA · Shanghai, China
I build visual SLAM and localization systems for autonomous vehicles — visual-inertial odometry, HD-map fusion, and robust state estimation. My working experience spans simulation, localization & mapping, environment information perception & fusion, and reinforcement learning.
Selected Research
Publications & Patents
A Vector Map Fusion Method Based on Cylindrical Point-Cloud Representation Combined with a Spatio-Temporal Attention Mechanism
End-to-end vector HD-map fusion from cylindrical point-cloud encodings with a spatio-temporal attention mechanism.
PaperCoarse-to-fine Semantic Localization with HD Map for Autonomous Driving In Structural Scenes
Camera-only vehicle localization that maps visual semantics to HD-map landmarks, refined by sliding-window pose-graph optimization.
PaperDirect Sparse Visual-Inertial Odometry with Stereo Cameras
Tightly-coupled stereo visual-inertial odometry built on direct, sparse photometric optimization.
Engineering
Selected Projects
Apollo × Unreal — Driving Simulation
Autonomous-driving simulation platform bridging the Baidu Apollo stack and the UISEE driving system inside Unreal Engine.
CalibrationSurround-Fisheye Online Extrinsic Calibration
Online extrinsic self-calibration for a surround-view fisheye rig using photometric consistency — the direct method from visual SLAM.
Field Notes