Quadrotor

I have done many projects about the location, mapping, planning and control for quadrotors on a low-performance chip. These projects are to make the quadrotor fly and explore in a unknown cluttered environment. Some demonstrations are listed as follows.

Real-time location, planning and control in unknown outdoor and indoor environment

  This research is to allow the quadrotor to fly in a unknown environments with only on-board sensors. A visual inertial odomery is used for location, improved from Vins-Fusion. MINCO is used to parameterize the trajectory while a L-BFGS optimizer is to solve the trajectory optimization in both spatial and temporal dimension. Then a SE3 controller based on inverse differential flatness mapping for quadrotor is used for low-level control.

Agressive crossing a narrow gap for quadrotor

  Construct a safe flight corridor (SFC) consisting of many convex polytopes for SE3 planning. Finely tune a SE3 controller with thrust calibration to accurately tracking the large-attitude trajectory for multicopter.

High-prediction stable hovering for quadrotor with a visual inertial odomery

  Design a new compact and light-weight structure for quadrotor. Optimize the VIO algorithm for low latency and high prediction for hovering.

Robot-centic ESDF and Integrated planning and control

  This video show the comparison between ego-planning-v2 and our planner. Due to the loss of a front-end guiding, the ego-planner tend to plan a long and slow trajectory.