This page is located in archive.

Semestral work - Frontier-based Exploration

The purpose of the semestral work is to create an environment in which the students will employ the knowledge gained and the implementations created during previous labs. Due to the current situation, the semestral work will be carried out individually (no teams), using a simulator instead of the real robots in the labs.

See instructions and source codes for details.

More details on the evaluation procedure and the final deadline will be announced later.

Evaluation details

For the evaluation purposes, we will distinguish only two classes in the occupancy grid: “Empty” (with occupancy confidence <0,25)) and “Other”, which merges Unknown (occupancy confidence -1) and Occupied (occupancy confidence <25,100> ). Given a ground truth occupancy grid and a predicted occupancy grid, the number of points will be determined according to the following table:

Predicted “Empty” <0,25) Predicted “Other” {-1, <25,100>}
Ground truth “Empty” 1 0
Ground truth “Other” -1 0

You will get positive points for correctly classified “Empty” voxels and negative points for incorrectly classified “Other” voxels.

In order to capture the temporal progress of your exploration, we will evaluate the “points” in the uniformly distributed temporal intervals: 30sec, 60sec, 90sec, 120sec, 180sec, and estimate “average_points” = 1/5 * (points_30 + points_60 + points_90 + points_120 + points_180).

The number of points from the semestral work (maximum is 22) will be determined as a saturated linear function of the “average_points” summed overall evaluation maps.


The deadline for uploading all codes to BRUTE is Thursday 4.6. 2020

courses/aro/tutorials/semestral_work.txt · Last modified: 2020/05/27 13:35 by zimmerk