uwreact / uwreact_robot

Software behind our fully autonomous FIRST robots

BSD 3-Clause "New" or "Revised" License

8 stars 6 forks source link

Localization around hab #21

Open wraftus opened 5 years ago

wraftus commented 5 years ago

🚀 Feature Request

We need to have a very accurate idea of where we are relative to the hab

matthew-reynolds commented 5 years ago

What do we plan on doing for special localization around the hab? I understand for most of the other specific localization issues, we're looking to leverage the alignment gaffers tape, the retroreflective targets, etc. What is special about the hab that needs/allows extra care beyond our standard global localization?

wmmc88 commented 5 years ago

As with the other local localization tasks, the goal is to eliminate any uncertainties or inaccuracies of the robots position by relying on local features rather than our amcl data.

In this case, goal is to line up with the center of the level 3 platform for our climb. This will be especially important if we're climbing with other robots.

Its open-ended how we achieve this. Maybe some distance measurement to wall using zeds or another time of flight sensor, maybe something to measure proximity to the HDPE or an array of sensors on the left and right ends of the front of our robot (that rests against the have wall), etc. Any sort of information that could help controls center the robots against the HDPE panel at the front of the level 3 climb platform

cyamonide commented 5 years ago

Potential data sources for localization:

Time of flight sensors (as mentioned above)
- Can measure distance from:
  - Front face of HAB
  - ~~PLAYER STATION 2 polycarbonate~~
- Reference points depend on sensor position on robot
Vision
- ~~Retroreflective tape~~
  - ~~Opposite of the HAB are two CARGO SHIP HATCHes facing it~~
  - Each HATCH has two vision targets (see Figure 4-32, section 4.10)
  - ~~These targets may be blocked with a HATCH PANEL~~
- Color/feature recognition
  - Each PLAYER STATION seems to have black tape as dividers _(see Figure 4-27, section 4.8.1.1)
  - HAB edge detection
  - Aforementioned CARGO SHIP HATCHes have white gaffers tape directly below them (see Figures 4-16 & 4-17, section 4.4)

wmmc88 commented 5 years ago

Depending on the ToF sensor, it could be buggy for it to work on the polycarbonate. But a well placed array of ToF could help us get a accurate idea of pose.

cyamonide commented 5 years ago

Sides of hab are black. This contrasts the alliance-coloured tops, and should make image processing easier.

Although I was unsure of hab colouring earlier, the field tour video and a Week 0 event seem to suggest the black sides.