1 of 11

If you need to be rescued, this is for you!

ROBOSAPIENS

André Gomes

Erya Guo

Maria Lopes

Oliver Hausdorfer

Precious Philip-Ifabiyi

Stefano Bassino

​

Team 4

2 of 11

SLAMMING AROUND

IMU-ODOMETRY + LIDAR POINTCLOUD

​

SLAM VERY WELL TUNED

​

SLAM

COMPUTATIONAL PROBLEMS ON LONGER EXPLORATIONS > ~ 5 MIN

3 of 11

EXPLORATION

HOW DID WE TUNE THE PARAMETERS?

4 of 11

EXPLORATION

HOW DID WE TUNE THE PARAMETERS?

5 of 11

HOW DO WE

GUARANTEE ALL OBJECTS ARE SEEN DURING CAMERA EXPLORATION?

​

​

WITH THE NORMAL EXPLORATION IT IS IMPOSSIBLE

6 of 11

REDUCING POINT CLOUD FOV!

(DIDN‘T WORK BECAUSE OF THE CURRENT TARE PLANNER IMPLEMENTATION)

7 of 11

OBJECT DETECTION

/object_detector/detection_info

/tf - get transform

Save timestamp, object type, and position in a .csv file

8 of 11

DETECTION POST-PROCESSING

DBSCAN

(Density-Based Spatial Clustering of Applications with Noise)

​

    • Clustering based on number of samples and inter-object distance.
    • Prevents multiple detection of the same object and averages their positions.

​

9 of 11

"Let me just change one thing ...”

10 of 11

11 of 11

QUESTIONS?

Thank you!