Jiawei Hou 侯佳维

Hou Jiawei graduated from Southwest Jiaotong University with a bachelor degree in Automation. After her graduation she joined the ShanghaiTech Automation and Robotics Lab as a master student. She brings great interest to work in the lab, hopes that she can be strong enough to one day to help the animals.


  • 2016.9 - Now:  ShanghaiTech University; Mobile Robotics PhD Candidate (2020.6)
  • 2012.9 - 2016.6: Southwest Jiaotong University Engineer Bachelor


  • Monocular Visual Odometry on the GPU

    In this project I estimated the camera pose transformation between images taken by a single camera at different points in time.

  • Full Speed Jackal

    We applied two ROS packages on a robot, Clearpath Jackal, driving by itself along a sub-optimal trajectory to a given goal point while avoiding obstacles.

    Team member: Jiawei Hou, Zhiming Li, Yanlin Zha

  • Participated in RoboCup Resce China 2017

    My main job is building a urdf file for our rescue robot.

  • The Area Graph -- A map topological representation

    We propose a novel map representation, the Area Graph, which makes the computation of large map much faster and easier. Areas of the Area Graph are generated from the Topology Graph to segment a given 2D map. The areas are connected by passages, which are usually doors and junctions, and contain Voronoi edges as paths through areas. 

  • Application to Path Planning Based on the Area Graph

    Based on the Area Graph, we build the Passage Graph for topological graph path planning in a map. The Passage Graph is built with the passages from the Area Graph as nodes and paths between passages as edges. Given a start and a goal, they are connected to their closest passages and the path between these two passages can be found by some simple search algorithm, e.g. the A* algorithm.

  • Map Matching

    Given 2 maps scanned in different sessions of the same environment with enough overlap, we match them by matching the corresponding areas of their Area Graphs. Then the transformation between 2 maps can be estimated from the correctly matched area pairs.