wh200720041/ssl_slam3

SSL_LIO: Solid-State LiDAR based LiDAR-Inertial SLAM

Wang Han [homepage]              

• :🌱: currently working in Huawei Singapore (Todmind Program).
• 🎓 received Ph.D. in Nanyang Technological University.
• 🔨 specialize in computer vision and robotics.

ssl_slam3's Issues

not enough correct surf points

Hello @wh200720041,

Thank you for the great work.

I setup the repo on my PC (ubuntu 20.04, with noetic) and tried the ssl_slam3.launch with the standard bag file and it worked fine. When I launch the ssl_slam3_L515.launch for the live run though, I immediately get the messages "not enough correct surf points" and "not enough correct edge points" and the map spirals out. I tried with all frequencies and resolutions (for RGB and depth) and the issue is still the same.

Hoping you could assist with the same.

IMU bias

Hi, @wh200720041

I have test the code for a modified version of traditional spinning lidar.
The only modification is feature extraction, the moudle is implemented based on lio-sam.
After the adjustment, everything is ok ,and it seems normal with my data.

I added a distortion correction module, use the imu preintegrate predict pose. But it seems something wrong, distortion correction did not work well.
I debug the question,and find out that the imu bias is small. It is not normal.

The possible reason maybe：

1. the adjustment is not right.
2. something wrong with the the imu preintegrate.

Can you give me some advice?
Thanks.

Derivation of jacobian of LidarEdgeFactor

hello, @wh200720041

Thanks for your great work.

I have a question about the derivation of jacobian formula in lidarOptimizationFactor.cpp

In factor LidarEdgeFactor:
Eigen::Matrix<double, 3, 15> dp_by_so3xyz = Eigen::Matrix<double, 3, 15>::Zero();
dp_by_so3xyz.block<3, 3>(0, 0) = - Ri * Utils::skew(curr_point);
dp_by_so3xyz.block<3, 3>(0, 3) = Eigen::Matrix3d::Identity();

I'm confused about why is a Ri here? how do you derive the formula?

Can you give me some help ?

Thanks

why "con.wait(lk, [&]{return synchronizeSensors();}); " last fo song long time.

I tested it, it lasted for about 40ms. Dose this part really need so much time?
And does your research group accept students from CSC exchange? I want to come and study, your work really impressed me.
Hope for your reply.

Derivation of jacobian formula

Hi, @wh200720041

Thanks for your great work.

I have a question about the derivation of jacobian formula in imuOptimizationFactor.cpp.

The residual function is:
`

// residual 1 = log[delta_r' * Ri' * Rj]
// residual 2 = Ri'* (0.5 * g *dt2 + Pj - Pi -Vi * dt) - delta_p
// residual 3 = Ri'* (g * dt + Vj - Vi) - delta_v
// residual 4 = b_aj - b_ai
// residual 5 = b_gj - b_gi

`

the optimized variable is r p v ba bg

I'm confused about how do you derive the formula
`

       jacobian_i.block<3, 3>(0, 0) = - Utils::Jr_so3_inv(rot_residual) * Rj.transpose() * Ri;
        jacobian_i.block<3, 3>(0, 12) = - Utils::Jr_so3_inv(rot_residual) * Rj.transpose() * Ri * corrected_delta_R * dr_dbg;

`

and jacobian_j.block<3, 3>(0, 0) = Utils::Jr_so3_inv(rot_residual);

Can you give me some help?

rslidar M1

I have a solid lidar called rslidar M1,could this code map successful on rslidar M1?