2024-01-08 Autonomy Leads Meeting
- 1 Agenda
- 2 Discussion
- 2.1 Bringing up new leads
- 2.2 Path going forward for current leads
- 2.3 Ongoing Projects
- 2.3.1 Pathing
- 2.3.2 Autonomy Airside
- 2.3.3 LTE Communication
- 2.3.4 IMACS
- 2.3.5 Machine Learning Model
- 2.4 Winter 2024 Administrative Tasks
- 2.5 Finance and Purchase
Agenda
Bringing up new leads
Path going forward for current leads
Ongoing projects
Winter 2024 Administrative Tasks
Finance and Purchase
Discussion
Bringing up new leads
Send out a google form to see who is interested in becoming a lead next term
To be eligible for becoming a lead the candidate must have prior PM experience on the team and also have been on the team for at least 4 months
Should ask what stream the person is in as a question in the form
Plan is to take on 2 leads this term (one onsite and one offsite) and in Spring 2024 we can put out a poll for a third lead
This leaves the team with 3 leads which is what we want as the team grows larger
The 2 leads will take on a lead/PM role
They learn to be a lead as they mentor a new PM into their respective projects
The people who have been selected to be a lead will be assigned the task to find a new PM for their project
Path going forward for current leads
Amy
Staying on as a member of the Autonomy team
Also thinking about joining the electrical subteam
Mihir
Staying on as a member of the Autonomy team
I want to take more of a mentorship role and continue to build more senior developers on the team
Also thinking about joining the EFS team (or try out another subteam)
Ongoing Projects
Pathing
Really good that Task 2 code is ready
Would like to have Task 2 code flight tested early in the term (TODO: Add to flight test schedule for this term)
Need to ground test Task 2 code with Pegasus to ensure waypoints can be uploaded
No work has been done on Task 1 so far
Task 1 has a lot of points for Autonomy at competition so this term we want to put high priority on getting Task 1 code done and ready for competition
@Mihir Gupta to coordinate with PM to do our best to get a plan out and people coding
Task 1 more difficult than Task 2 because there are a lot more variables to account for
Autonomy Airside
Detect target task is blocking a lot of integration tasks
Need to get that done so we can move forward with Geolocation
@Mihir Gupta to pair program with person assigned to task to get this done as soon as possible
We should put strict deadlines on the blocking tasks and tell people it is very critical to complete on time
@Mihir Gupta would be happy if we got to a point where we have conducted an integration test with Geolocation (and hopefully have it working)
Geolocation is the backbone for many computer vision tasks in the AEAC and other drone competitions
Having a tested and working Geolocation pipeline would make it a lot easier to expand our Airside system for future competitions
Would be a huge win going forward for the team
@Amy Hu also wants to test Cluster Estimation as well as it is another key component (accounts for errors in ML model and geolocation algorithm)
A big hurdle we would like to work on this term is getting Geolocation working
LTE Communication
Need to focus on transmitting images to the Jetson (which will be on the ground)
We are able to transfer drone telemetry to the Jetson
@Mihir Gupta I did not do a good job scoping out how we should send data over with LTE
Did not know a lot about compressions techniques for images
Hopefully with more time during coop I can learn more about the compression techniques and start getting the repository ready
Also write documentation on how the LTE system will work at competition
Clarify when we will start transmitting data
We need to be able to ground test the system to ensure everything works as expected
Ground test should be enough to verify the system works
@Mihir Gupta thinking about asking Maxwell if he wants to be PM
Worth reaching out to him
IMACS
Progress is that data telemetry works and we are starting on the GUI tasks
PM is scoping out how the UI should look like and how the tasks can be done
Steady progress
Good project as it gets people on the team with tasks
If we didn’t have this project it would be difficult to assign everyone on our team a task as our team is growing a lot
Goal for this term is to make more progress on the GUI side and hopefully have a screen constructed
Â
Machine Learning Model
Data has been collected but no word on data cleaning as of now
Ask current PM if he wants to stay as PM as he is on exchange right now
@Mihir Gupta do some administrative work and get the tasks for data cleaning unblocked
Get the directories organized so people can be assigned them
Add more to the doc and make it more comprehensive
Hope to get data cleaning done soon so we know how many images we have
If we don’t have enough we may have to do some more flight tests for image collection
As a backup if we don’t have enough images @Mihir Gupta we can use the images from 2022 that Xierumeng collected and annotate those (unideal but if we are in that situation may be the best choice we have)
Winter 2024 Administrative Tasks
@Amy Hu
Cleaning up Asana
In-person room booking
@Mihir Gupta
Create Flight test schedule
Create a Project Plan document for Directors Sync
Both of us
Set a time for meetings
Wednesday at 7:00 PM to 8:00 PM
New Projects?
Both of us think we have enough and want to double down on competition projects
Future Project ideas
Trying to convert our machine model to a Tensorflow Lite model
This way we can run our model on Google TPU’s from Coral AI
Reduces cost since we only need RPi + TPU and improves performance as TPU’s run much faster than NVIDIA GPU’s
Combining LTE and Autonomy airside code together so it all runs on one board (maybe we can try to purchase an NVIDIA Jetson Nano)
Machine Learning with TOF to estimate distance (perform depth estimation)
Get Point Pillar data from a LiDAR and maybe construct a map of the scene (SLAM project)
Finance and Purchase
In Fall 2023 @Mihir Gupta presented to MEF
We got our first option in funding (which is $930.85)
We can spend it on the following items
Jetson carrier
CV Camera
Raspberry Pi’s
@Amy Hu will think about presenting this term
Autonomy lead at these meetings are helpful because they can discuss all the math that is used in our system
Makes our presentation more relevant to Math
We also got around $500 from WEEF too
We got this for LiDAR
We need to make some purchases quickly
New CV Camera @Amy Hu
Current CV Camera has broken inductor casing
Flight testing a lot so has high probability that it can break
Same model as the current one we have
Raspberry Pi @Mihir Gupta
LTE will need to do some testing with the Raspberry Pi
Also good to have backup as we only have 1 to use
Purchase the Raspberry Pi 4B
Make these purchases at the beginning of the term