Ground truthing CoPeD

https://wp.nyu.edu/arpl/

Abhimanyu Suthar

New York University

​

Context

• For the CoPeD dataset, there is a need to get reliable ground truth in order to improve annotations and establish a more robust benchmark for evaluating algorithms

​

• The dataset can be found here: https://drive.google.com/drive/folders/1ldGY2qRKa0tmDDnYDLmaHRV1wBKn7NAM

​

​

2

TASK

• I made use of LOCUS to get local lidar odometry and then fed those estimates into LAMP that consists of a base station and created a global map.

​

• The steps:
• Configure the rosbags to work with LOCUS system, by remapping the LiDAR, odometry and IMU topics to work with the wanda robot topics

​

2. After getting the lidar odometry result, fed this into the LAMP system and received the global map.

​

​

3

LOCUS -HOUSE:A

https://drive.google.com/file/d/1bWYV117e_dC1bvid3aPmb1m6BdVo3FV3/view?usp=sharing

4

LAMP - HOUSEA

5

Plots

6

HOUSE-A: Lidar odometry

7

HOUSE-A : Lidar odometry

8

Overlapping GPS, LO, rosbag

9

Comparing position

10

Comparing Roll Pitch Yaw

11

Context

• I also ran tests and analysed raw Lidar data directly from the rosbag using the topic $(ROBOT_NAME)/lidar_points

​

The above analysis is done using Open3d library

12

More information on analysis here

https://drive.google.com/drive/folders/1me3m_HzLEwKcekAAoLEZ70pESUEcs2dr?usp=share_link

13

Thank you

https://wp.nyu.edu/arpl/

Abhimanyu Suthar

New York University

abhimanyu.suthar@nyu.edu