---
title: Meeting 9/11/21
tags: meeting-notes
---
# Meeting 9/11/21

### General Action Items
- Fill out [Github form](https://forms.gle/b84vftCSDLFosxMa7) to get access to github
- Fill out the [whenisgood](http://whenisgood.net/eatdteg) form to share your availibility
- Join the software role in the #roles channel to get announcements and notifications regarding software stuff (feel free to mute other subteam categories to limit "spam notifs")
### Discussion
- [ ] Get to know the team and what they're interested in (If you're not on here, add yourself to this list please)
- High-level
- Add here
- Navigation
- Ashvin,Stuti, Alice
- Localization
- Evan,Stuti, Allison

- Task-level Control
- Ashvin
- Low-level
- Add here
- Interested in doing all
- Raghava, Naren
- [ ] Discuss some subareas more in-depth
- High-level
- Looking into mapping out all the states the robot could be in a behavior tree or finite state machine
- [Tutorial by robotic sea bass](https://roboticseabass.com/2021/05/08/introduction-to-behavior-trees/)
- Navigation
- Localizing craters + rocks
- Using a neural network and SOTA object detection pipelines to detect craters
- Pretrained models for transfer learning: YOLO, ImageNet, ResNet50
- How would we collect a dataset?
- Can our network generalize? One-shot learning because we cannot access image space "on the moon"
- Domain randomization?
- Use image processing algorithms that are robust enough to detect circular crater region (2-3 points of crater is enough to localize it with depth camera)
- Generating an optimal path around discovered obstacles (being able to update the path as new obstacles are localized)
- Spline generation?
- RRT? A-star?
- Going simple first with just following a tutorial
- [A-star with Turtlebot](https://github.com/aarjunsrinivasan/A-Star-using-TurtleBot-ROS)
- Uses ROS melodic, unsure if it works on noetic, but it should
- Localization
- BLE sensor
- Trying to get things working with phones + raspberry pi to minimize assumptions
- TODO: Need more accurate info on wifi/bluetooth drivers
- Use Jetson Xavier NX, maybe their datasheet has details?
- Task-level Control
- Pretty much fully dependent on exdep downselection
- Drivetrain needs encoders on their motors for velocity feedback + control (can use T265 localization to help out here as well)

### Future Discussion

#### General

- How is the wiki? Too complex? Not enough info? Too vague?
- BIDC access

#### High-level

#### Navigation

#### Localization

#### Task-level Control

#### Low-level