@zauberflote1 @Cliff-Wong @Alex-Kushleyev @Eric-Katzfey
Hey guys, I am using voxl-open-vins-server.
I facing an issue of constant drift, the drone is drifting while takeoff itself. Once I take off itself the vins values varies from 1 m to 80 m. I take off in stabilised for safety. The drone once changed to position mode, constantly moves in one direction, until pilot gives different command, again it will continue in the direction of the pilot command.
I am mentioning my config file here, please guide me to achieve strong position hold.
/**
* This file contains configuration that's specific to voxl-open-vins-server.
*
* *NOTE*: all time variables are measured in seconds
*
* OpenVins param breakdown:
*
* do_fej: whether or not to do first estimate Jacobians
* imu_avg: whether or not use imu message averaging
* use_rk4_integration: if we should use Rk4 imu integration.
* cam_to_imu_refinement: whether or not to refine the imu-to-camera pose
* cam_intrins_refinement: whether or not to refine camera intrinsics
* cam_imu_ts_refinement: whether or not to calibrate cam to IMU time offset
* max_clone_size: max clone size of sliding window
* max_slam_features: max number of estimated SLAM features
* max_slam_in_update: max number of SLAM features in a single EKF update
* max_msckf_in_update: max number of MSCKF features used at an image timestep
*
* Feature Reps can be any of the following:
* 0 - GLOBAL_3D
* 1 - GLOBAL_FULL_INVERSE_DEPTH
* 2 - ANCHORED_3D
* 3 - ANCHORED_FULL_INVERSE_DEPTH
* 4 - ANCHORED_MSCKF_INVERSE_DEPTH
* 5 - ANCHORED_INVERSE_DEPTH_SINGLE
* feat_rep_msckf: (int) what representation our msckf features are in
* feat_rep_slam: (int) what representation our slam features are in
* cam_imu_time_offset: time offset between camera and IMU
* slam_delay: delay that we should wait from init before estimating SLAM features
* gravity_mag: gravity magnitude in the global frame
* init_window_time: amount of time to initialize over
* init_imu_thresh: variance threshold on our accel to be classified as moving
*
* imu_sigma_w: gyroscope white noise (rad/s/sqrt(hz))
* imu_sigma_wb: gyroscope random walk (rad/s^2/sqrt(hz))
* imu_sigma_a: accelerometer white noise (m/s^2/sqrt(hz))
* imu_sigma_ab: accelerometer random walk (m/s^3/sqrt(hz))
* imu_sigma_w_2: gyroscope white noise covariance
* imu_sigma_wb_2: gyroscope random walk covariance
* imu_sigma_a_2: accelerometer white noise covariance
* imu_sigma_ab_2: accelerometer random walk covariance
*
* ****_chi2_multiplier: what chi-squared multipler we should apply
* ****_sigma_px: noise sigma for our raw pixel measurements
* ****_sigma_px_sq: covariance for our raw pixel measurements
* use_stereo: if feed_measurement_camera is called with more than one
* image, this determines behavior. if true, they are treated as a stereo
* pair, otherwise treated as binocular system
* if you enable a camera with stereo in the name, this will be set to true
* automatically
*
* try_zupt: if we should try to use zero velocity update
* zupt_max_velocity: max velocity we will consider to try to do a zupt
* zupt_only_at_beginning: if we should only use the zupt at the very beginning
* zupt_noise_multiplier: multiplier of our zupt measurement IMU noise matrix
* zupt_max_disparity: max disparity we will consider to try to do a zupt
* *NOTE*: set zupt_max_disparity to 0 for only imu based zupt, and
* zupt_chi2_multipler to 0 for only display based zupt
*
* num_pts: number of points we should extract and track in each image frame
* fast_threshold: fast extraction threshold
* grid_x: number of column-wise grids to do feature extraction in
* grid_y: number of row-wise grids to do feature extraction in
* min_px_dist: after doing KLT track will remove any features closer than this
* knn_ratio: KNN ration between top two descriptor matchers for good match
* downsample_cams: will half image resolution
* use_nultithreading: if we should use multi-threading for stereo matching
* use_mask: if we should load a mask and use it to reject invalid features
*/
{
"en_auto_reset": true,
"auto_reset_max_velocity": 20,
"auto_reset_max_v_cov_instant": 0.10000000149011612,
"auto_reset_max_v_cov": 0.10000000149011612,
"auto_reset_max_v_cov_timeout_s": 0.5,
"auto_reset_min_features": 1,
"auto_reset_min_feature_timeout_s": 3,
"auto_fallback_timeout_s": 3,
"auto_fallback_min_v": 0.600000023841858,
"en_cont_yaw_checks": false,
"fast_yaw_thresh": 5,
"fast_yaw_timeout_s": 1.75,
"do_fej": true,
"imu_avg": true,
"use_rk4_integration": true,
"cam_to_imu_refinement": true,
"cam_intrins_refinement": true,
"cam_imu_ts_refinement": true,
"max_clone_size": 8,
"max_slam_features": 35,
"max_slam_in_update": 10,
"max_msckf_in_update": 10,
"feat_rep_msckf": 4,
"feat_rep_slam": 4,
"cam_imu_time_offset": 0,
"slam_delay": 1,
"gravity_mag": 9.80665,
"init_window_time": 1,
"init_imu_thresh": 1,
"imu_sigma_w": 0.00013990944749616306,
"imu_sigma_wb": 4.1189724174615527e-07,
"imu_sigma_a": 0.0038947538150776763,
"imu_sigma_ab": 5.538346201712153e-05,
"msckf_chi2_multiplier": 1,
"slam_chi2_multiplier": 40,
"zupt_chi2_multiplier": 1,
"msckf_sigma_px": 4,
"slam_sigma_px": 4,
"zupt_sigma_px": 4,
"try_zupt": true,
"zupt_max_velocity": 0.02,
"zupt_only_at_beginning": true,
"zupt_noise_multiplier": 1.5,
"zupt_max_disparity": 8,
"init_dyn_use": true,
"triangulate_1d": false,
"refine_features": true,
"max_runs": 5,
"init_lamda": 0.001,
"max_lamda": 10000000000,
"min_dx": 1e-06,
"min_dcost": 1e-06,
"lam_mult": 10,
"min_dist": 0.1,
"max_dist": 60,
"max_baseline": 40,
"max_cond_number": 600000,
"use_mask": false,
"use_multithreading": true,
"use_stereo": true,
"use_baro": false,
"num_opencv_threads": 4,
"fast_threshold": 30,
"histogram_method": 1,
"knn_ratio": 0.7,
"takeoff_accel_threshold": 0.80,
"takeoff_threshold": 0.3,
"use_stats": false,
"max_allowable_cep": 1,
"en_force_init": false,
"en_force_ned_2_flu": false,
"track_frequency": 40,
"publish_frequency": 15,
"en_vio_always_on": false,
"en_ext_feature_tracker": false,
"en_gpu_for_tracking": true,
"num_features_to_track": 40
}
