Home

This stack contains a sensor fusion framework based on an Extended Kalman Filter (EKF) for vehicle pose estimation including intra- and inter-sensor calibration. It assumes full 6DoF motion of the vehicle and an IMU centred platform. It is a self-calibrating approach rendering a vehicle a true power-on-and-go system. More precisely it estimates the vehicle 6DoF pose and velocity which can be used in an underlying pose controller IMU biases such as acceleration and gyroscope bias (intra-sensor calibration) sensor transformation between an additional sensor (e.g. camera, GPS) and the IMU (inter-sensor calibration) pose scaling in the case of a scaled pose measurement sensor (e.g. camera) roll and pitch drift of a non-global pose measurement sensor (e.g. camera). The framework is particularly designed to work on an Micro Aerial Vehicle (MAV) carrying an IMU and one single camera performing visual odometry as only navigation sensors (see publications below and ethzasl_ptam). It has the following properties:

The pose and velocity estimates yield sufficiently clean values for pose control the MAV with this filter output. If used together with the asctec_mav_framework and ethzasl_ptam stacks reliable pose estimates at 1kHz are possible. The intra- and inter-sensor calibration estimates make it possible to deploy this algorithm quickly on any vehicle without tedious calibration procedures. Thus, this framework renders your platform power-on-and-go in real environments. The scaling and drift estimates ensure long term operability despite potential drift of an additional sensor such as the visial odometry when using a monocular camera. Note that only the roll and pitch drift of the visual odometry are observable, however, these are the most crucial states to keep an MAV airborne as it ensures the alignment to gravity.