search

baidu / boteye

211 stars 72 forks source link

Glad to know this algorithm but encountered with confusions. #9

Closed TiderFang closed 6 years ago

TiderFang commented 6 years ago

I happened to know this SLAM algorithm and I am quite curious about it. But it's disappointing that I couldn't find ever any videos or implements of this SLAM algorithm describing its greatness as what the hector slam, cartographer and gmapping do. Also there is no article related to it. As an open source SLAM algorithm, that's strange. Could you give me some tips for finding these treasure ? I have read the Tech Specs in the introduction page. Several questions appear as below:

  1. SLAM laterncy < 10 us? Hard to believe. If its true, so it's a realtime algorithm. (Here I regard the SLAM laterncy as the map update rate).
  2. Also I have no idea with the SLAM error concept. Could anyone answer my confusions about this algorithm? Best wishes.
mingyux commented 6 years ago

@TiderFang We're working hard to open source the core of our SLAM engine. As for your questions:

  1. The pose rate is identical to the IMU rate, which is 100 Hz for our current sensor. The optimization rate follows the image rate, which is 20 Hz.
  2. To evaluate the performance of a odometry-based algorithm, we use the drift error percentage, i.e., the L2 distance error / the total travel distance.

We also have a technical paper describing our SLAM system, which is still under review. Please stay tuned.

TiderFang commented 6 years ago

I see. Thanks

kiyology commented 6 years ago

Closing this as it's inactive for 3 weeks.