Project: Perception Pick & Place

1. Provide a Writeup / README that includes all the rubric points and how you addressed each one. You can submit your writeup as markdown or pdf.

You're reading it!

Exercise 1, 2 and 3 pipeline implemented

1. Complete Exercise 1 steps. Pipeline for filtering and RANSAC plane fitting implemented.

Here is a link to my python script for all exercises & the project.

My pipeline is as follows:

Image -> Voxel Grid Downsampling -> Passthrough over Z-Axis -> Passthrough over X-Axis -> Outlier Filter -> RANSAC PLANE Filter -> Done

For each of these sections, I'll list my observactions:

Voxel Grid: The hardest thing to learn here was what the LEAF_SIZE actually meant. I kept thinking it meant that something like "1" meant one point in each "area." I didn't realize that the leaf is actually defining the "area."

Passthrough Filters: The biggest killer to me on this one was remembering the right hand rule when I was learning it. This one is pretty easy, but will be my biggest challenge when I take the challenge_world.

Outlier Filter: After the working python_pcl was given to us, this one is so simple that it isn't worth mentioning.

RANSAC PLANE FILTER: This one still perplexes me. I understand the concept, just not really how it works, I need to mess with this one some more.

Here is what the cloud looked like after the pipeline:

2. Complete Exercise 2 steps: Pipeline including clustering for segmentation implemented.

I used the same Euclidean Clutering that was given to us a sample. For the project as opposed to the exercises, I had to mess around with the values a lot more until I got things that worked.

Here is what the cloud looked like after the pipeline:

3. Complete Exercise 3 Steps. Features extracted and SVM trained. Object recognition implemented.

Here are the settings I used to create my SVM:

Pictures: 500

Bins (Color & Normals): 64

RGB vs HST: HST

Total Features: 4000

Here's the output of the training:

Here is the object and their markers:

Pick and Place Setup

1. For all three tabletop setups (test*.world), perform object recognition, then read in respective pick list (pick_list_*.yaml). Next construct the messages that would comprise a valid PickPlace request output them to .yaml format.

First, the results:

World 1 100%:

YAML File: Link!

World 2 100%:

YAML File: Link!

World 3 100%:

YAML File: Link!

Now, my thoughts:

The exercises took us 80% of the way, which was a nice way of handling this project. Taking my exercise code and moving it to the project was easy. Getting the yaml file output was fun and easy also. The only major challenge I found was upping the number of features from 200 to 4000 in the training. This got me 100% of objects in all worlds.