Vins-Mono / Vins-Fusion usage

[aimiliosnrobotics]

Hello and thank you very much for the nice work! This question is about the usage of the vins-mono and vins-fusion packages with my own data. I'm facing a problem that i cant use the vins-fusion project. I am running successfully the vins-mono with my setup (custom monocular camera + phidget spatial IMU) and I am able to get correct results but when running the vins-fusion with the exact same configurations as the vins mono the camera_pose drifts heavily away as it can be seen in the following images.

VINS MONO:

VINS FUSION:

Has something changed between the packages ? Its also worth mentioning, that for the vins mono i had to inflate the imu noise parameters in order to make it run successfully, although i got the imu parameters with alan variance analysis and was able to successfully get very accurate results using these parameters with other VI-slam packages like orb-slam. Is the vins package more sensitive to bad imu measurements or exclusively planar motion with approximately constant acceleration ? My ultimate goal is to use VI-multicamera slam for a wheeled robot and the whole package works well with orb slam 3 but have some cases that there is drifting and wrong scaling and after doing a lot of tests I thought it was because of the planar setup and approximately constant acceleration. Thats why im trying to ultimately use the vins-gps-wheel package to fuse wheel odometry to recover the missing scale information and optionally gps. I am running vins-mono almost successfully (apart from the inflated imu noises) so i think the vins-wheel odometry coupling would be successful as the package is based on vins-mono (already can see the feature pointcloud published. However, im trying the vins-fusion package because it is said that it supports multiple cameras, so maybe i could use it to finally have the multicamera-VI slam - wheel odometry fusion. So does anyone have any insight on this and why the Vins-fusion package is not running ? Thank you very much in advance!

the configuration files are following:

VINS-MONO:

%YAML:1.0

#common parameters
#support: 1 imu 1 cam; 1 imu 2 cam: 2 cam; 
imu_topic: "/imu/data_raw"
image_topic: "/camera0/image_raw"
#image1_topic: "/camera1/image_raw"
output_path: "~/output/"

# model_type: KANNALA_BRANDT
# camera_name: camera0
# image_width: 640
# image_height: 480
# distortion_parameters:
#    k1: 0.40000818354392403
#    k2: -0.12081171870317779
#    p1: -0.05019646239461993
#    p2: 0.03191883340304046
# projection_parameters:
#    fx: 249.11357612128313
#    fy: 249.11690732189138
#    cx: 326.68403379291203
#    cy: 243.31467455420616   

model_type: PINHOLE
camera_name: camera0
image_width: 640
image_height: 480
distortion_parameters:
   k1: -0.06155816710653534
   k2: -0.013961160967415768
   p1: 0.0007546126094149863
   p2: -0.0008787572974087528
projection_parameters:
   fx: 258.51567840106213
   fy: 258.5752935695983
   cx: 321.1744166775173
   cy: 242.3615494668013

# Extrinsic parameter between IMU and Camera.
estimate_extrinsic: 0   # 0  Have an accurate extrinsic parameters. We will trust the following imu^R_cam, imu^T_cam, don't change it.
                        # 1  Have an initial guess about extrinsic parameters. We will optimize around your initial guess.

#### KANNALABRANDT #############
# extrinsicRotation: !!opencv-matrix
#    rows: 3
#    cols: 3
#    dt: d
#    data: [-0.99926158, -0.02532093, -0.02889875, 
#          -0.02768236, -0.04713959, 0.99850466, 
#          -0.02664534, 0.99856733, 0.04640383]
# #Translation from camera frame to imu frame, imu^T_cam
# extrinsicTranslation: !!opencv-matrix
#    rows: 3
#    cols: 1
#    dt: d
#    data: [0.04549169,0.31585158, 0.03692635]

#### PINHOLE #############
extrinsicRotation: !!opencv-matrix
   rows: 3
   cols: 3
   dt: d
   data: [-0.99948037, -0.01843143, -0.0264439, 
          -0.02545083, -0.0521769, 0.99831349,
          -0.01978011, 0.99846775, 0.0516807]
#Translation from camera frame to imu frame, imu^T_cam
extrinsicTranslation: !!opencv-matrix
   rows: 3
   cols: 1
   dt: d
   data: [0.04378348,0.32805217, 0.03791052]

#feature traker paprameters
max_cnt: 150            # max feature number in feature tracking
min_dist: 30            # min distance between two features 
freq: 10                # frequence (Hz) of publish tracking result. At least 10Hz for good estimation. If set 0, the frequence will be same as raw image 
F_threshold: 1.0        # ransac threshold (pixel)
show_track: 1           # publish tracking image as topic
equalize: 1             # if image is too dark or light, trun on equalize to find enough features
fisheye: 0              # if using fisheye, trun on it. A circle mask will be loaded to remove edge noisy points

#optimization parameters
max_solver_time: 0.04  # max solver itration time (ms), to guarantee real time
max_num_iterations: 8   # max solver itrations, to guarantee real time
keyframe_parallax: 10.0 # keyframe selection threshold (pixel)

# #imu parameters       The more accurate parameters you provide, the better performance
# acc_n: 0.0032171141081418185          # accelerometer measurement noise standard deviation. 
# gyr_n: 0.00047514840397045084         # gyroscope measurement noise standard deviation.     
# acc_w: 0.002008277168849152        # accelerometer bias random work noise standard deviation.  
# gyr_w: 1.4322615524203976e-05       # gyroscope bias random work noise standard deviation.     
# g_norm: 9.81007     # gravity magnitude

acc_n: 0.08          # accelerometer measurement noise standard deviation. #0.2   0.04
gyr_n: 0.004         # gyroscope measurement noise standard deviation.     #0.05  0.004
acc_w: 0.00004         # accelerometer bias random work noise standard deviation.  #0.02
gyr_w: 2.0e-6       # gyroscope bias random work noise standard deviation.     #4.0e-5
g_norm: 9.81007     # gravity magnitude



#loop closure parameters
loop_closure: 1                    # start loop closure
load_previous_pose_graph: 0        # load and reuse previous pose graph; load from 'pose_graph_save_path'
fast_relocalization: 0             # useful in real-time and large project
pose_graph_save_path: "/home/shaozu/output/pose_graph/" # save and load path

#unsynchronization parameters
estimate_td: 0                      # online estimate time offset between camera and imu
td: 0                             # initial value of time offset. unit: s. readed image clock + td = real image clock (IMU clock)

#rolling shutter parameters
rolling_shutter: 0                  # 0: global shutter camera, 1: rolling shutter camera
rolling_shutter_tr: 0               # unit: s. rolling shutter read out time per frame (from data sheet). 

#visualization parameters
save_image: 1                   # save image in pose graph for visualization prupose; you can close this function by setting 0 
visualize_imu_forward: 0        # output imu forward propogation to achieve low latency and high frequence results
visualize_camera_size: 0.4      # size of camera marker in RVIZ

VINS-FUSION:

%YAML:1.0

#common parameters
#support: 1 imu 1 cam; 1 imu 2 cam: 2 cam; 
imu: 1         
num_of_cam: 1  

imu_topic: "/imu/data_raw"
image0_topic: "/camera0/image_raw"
#image1_topic: "/camera1/image_raw"
output_path: "~/output/"

cam0_calib: "cam0.yaml"
#cam1_calib: "cam1.yaml"
image_width: 640
image_height: 480
   

# Extrinsic parameter between IMU and Camera.
estimate_extrinsic: 0   # 0  Have an accurate extrinsic parameters. We will trust the following imu^R_cam, imu^T_cam, don't change it.
                        # 1  Have an initial guess about extrinsic parameters. We will optimize around your initial guess.

#### KANNALABRANDT #############
# body_T_cam0: !!opencv-matrix
#    rows: 4
#    cols: 4
#    dt: d
#    data: [-0.99926158, -0.02532093, -0.02889875, 0.04549169, 
#           -0.02768236, -0.04713959, 0.99850466, 0.31585158,
#           -0.02664534, 0.99856733, 0.04640383, 0.03692635,
#           0,0,0,1]

#### PINHOLE #############
body_T_cam0: !!opencv-matrix
   rows: 4
   cols: 4
   dt: d
   data: [-0.99948037, -0.01843143, -0.0264439, 0.04378348,
          -0.02545083, -0.0521769, 0.99831349, 0.32805217,
          -0.01978011, 0.99846775, 0.0516807, 0.03791052,
          0,0,0,1]

#Multiple thread support
multiple_thread: 1

#feature traker paprameters
max_cnt: 150            # max feature number in feature tracking
min_dist: 30            # min distance between two features 
freq: 10                # frequence (Hz) of publish tracking result. At least 10Hz for good estimation. If set 0, the frequence will be same as raw image 
F_threshold: 1.0        # ransac threshold (pixel)
show_track: 1           # publish tracking image as topic
flow_back: 1            # perform forward and backward optical flow to improve feature tracking accuracy

#optimization parameters
max_solver_time: 0.04  # max solver itration time (ms), to guarantee real time
max_num_iterations: 8   # max solver itrations, to guarantee real time
keyframe_parallax: 10.0 # keyframe selection threshold (pixel)

#imu parameters       The more accurate parameters you provide, the better performance
# acc_n: 0.0032171141081418185         # accelerometer measurement noise standard deviation. 
# gyr_n: 0.00047514840397045084         # gyroscope measurement noise standard deviation.     
# acc_w: 0.002008277168849152        # accelerometer bias random work noise standard deviation.  
# gyr_w: 1.4322615524203976e-05       # gyroscope bias random work noise standard deviation.     
# g_norm: 9.81007     # gravity magnitude

acc_n: 0.08          # accelerometer measurement noise standard deviation. #0.2   0.04
gyr_n: 0.004         # gyroscope measurement noise standard deviation.     #0.05  0.004
acc_w: 0.00004         # accelerometer bias random work noise standard deviation.  #0.02
gyr_w: 2.0e-6       # gyroscope bias random work noise standard deviation.     #4.0e-5
g_norm: 9.81007     # gravity magnitude

# acc_n: 0.1          # accelerometer measurement noise standard deviation. 
# gyr_n: 0.01         # gyroscope measurement noise standard deviation.     
# acc_w: 0.001        # accelerometer bias random work noise standard deviation.  
# gyr_w: 0.0001       # gyroscope bias random work noise standard deviation.     
# g_norm: 9.81007     # gravity magnitude
#unsynchronization parameters
estimate_td: 0                      # online estimate time offset between camera and imu
td: 0                             # initial value of time offset. unit: s. readed image clock + td = real image clock (IMU clock)

#loop closure parameters
load_previous_pose_graph: 0        # load and reuse previous pose graph; load from 'pose_graph_save_path'
pose_graph_save_path: "~/output/pose_graph/" # save and load path
save_image: 1                   # save image in pose graph for visualization prupose; you can close this function by setting 0 

cam0.yaml

%YAML:1.0
---
model_type: PINHOLE
camera_name: camera
image_width: 640
image_height: 480
distortion_parameters:
   k1: -0.06155816710653534
   k2: -0.013961160967415768
   p1: 0.0007546126094149863
   p2: -0.0008787572974087528
projection_parameters:
   fx: 258.51567840106213
   fy: 258.5752935695983
   cx: 321.1744166775173
   cy: 242.3615494668013

# model_type: KANNALA_BRANDT
# camera_name: camera0
# image_width: 640
# image_height: 480
# distortion_parameters:
#    k1: 0.40000818354392403
#    k2: -0.12081171870317779
#    p1: -0.05019646239461993
#    p2: 0.03191883340304046
# projection_parameters:
#    fx: 249.11357612128313
#    fy: 249.11690732189138
#    cx: 326.68403379291203
#    cy: 243.31467455420616

Best regards
Aimilios Petroulas

[mansisarawata]

Hi @aimiliosnrobotics were you able to solve this?