Month: March 2024

Team Status Report for 3/16/24

Team Status Report for 3/16/24

Potential Risks and Mitigation Strategies The biggest thing we are uncertain about right now is whether or not the videos we are taking at the actual intersection we want to model (Fifth and Craig) will be sufficient to train the model. It could be challenging 

Kaitlyn’s Status Report for 3/16/24

Kaitlyn’s Status Report for 3/16/24

Work Done This week I spent a lot of time working on the Ethics Assignment and also finishing up the simulation code to use in our ML model. At the moment, I am able to remotely run a SUMO simulation in Python using TraCI. I 

Zina’s Status Report for 3/9/24

Zina’s Status Report for 3/9/24

Given that the Design Report was due this week, we had to lock in a lot of the details that we were uncertain about up until now. The process of writing the report was very helpful and made us think critically about the more challenging parts of our project. Up until this week, we had assumed that we would be using addressable LEDs to represent our traffic light mockup circuit, but when I was doing research on this type of LED I realized that most available ones can draw up to 60 mA of current per LED. Since we will be using an Arduino to control the light timings, and an Arduino’s digital GPIO pins can only output a maximum of 40 mA of current (and 20 mA of constant current) without causing damage to the hardware, this was not going to be a feasible solution. Therefore I made the design decision to instead have the Arduino drive a Texas Instruments TLC5928 LED Driver module which can output a constant current to up to 16 different LEDs by setting a reference current by tying an apt resistor to the module’s V_CC pin. This will allow us to draw power from the Arduino’s 5V output pin, which can supply up to 900mA of current, instead of the individual GPIO pins. The GPIO pins will instead drive the logic at the input pins of the LED Driver module. All 12 LEDs that will represent our traffic light circuit as well as the LED Driver module will be soldered to our custom breakout PCB for the Arduino so that our physical intersection simulation can all be integrated into one place. The schematic mockup I created in Eeschema for this circuit is below.

My plan for the next week is to get the necessary parts ordered and learn the protocol for the TLC5928 unit’s input logic so that I can write Arduino code to drive our LEDs. I would also ideally like to have our PCB layout completed, but we will wait to order it until the LED Driver module comes in so that I can test that everything is working as intended before we get our PCBs printed.

Team Status Report for 3/9/24

Team Status Report for 3/9/24

Potential Risks and Mitigation Strategies Currently we are having challenges setting up the RPi to work with the CMU-Secure WiFi. In case that does not work out, we will instead use a mobile hotspot to provide internet access to our RPi to perform the necessary 

Kaitlyn’s Status Report for 3/9/24

Kaitlyn’s Status Report for 3/9/24

Work Done I spent most of the week working on the Design Report and focused on the introduction, project management, and content relating to the optimization algorithm. There were actually a lot of details that we had to flesh out relating to the optimization algorithm 

Ankita’s Status Report for 3/9/24

Ankita’s Status Report for 3/9/24

Work Done

Last week, I along with our group members worked on and completed our 12-page design report. I completed the design requirements, block diagrams, and summary, as well as the architectural, implementation, testing, and trade study descriptions of the camera interfacing with the Raspberry Pi and the object detection algorithm. To write the design report we had to flesh out many aspects of our project; we have decided to connect the RPi and IP cameras to a Mobile WiFi hotspot (I have an unlimited data plan, so cost will not be an issue). Connecting the RPi to CMU-DEVICE with a headless setup has proved to be difficult; I registered the MAC address with IT services but was unfortunately not able to ssh into the Pi’s network.

Schedule

Working on the design report took most of my time this week so I wasn’t able to train the new Haar classifier, though I have a better idea on how to do it now. I need some “negative” images to tag in order to train the classifier (images of intersections with 0 cars) so once I get back to Pittsburgh Zina and I will work on getting those taken. Because of this, I will probably simplify the pedestrian detection model to a simple “yes/no” classifier (as in, are there pedestrians waiting to cross?) instead of counting exactly how many pedestrians are on each side (since pedestrians tend to walk side by side when they cross anyway.) That way we can just use a pre-trained classifier or limit the accuracy of the classifier we do have.

Deliverables

By the end of next week, I will:

  • Train the new vehicle classifier.
  • Connect the RPi and the IP Cameras to the Mobile WiFi hotspot.
  • Figure out what I need to do to capture an image with the BLE camera.