Team Status Report for 3/30/24

Potential Risks and Mitigation Strategies

The new YOLO car detection model is working a lot better and way more accurate than our previous model with the Haar cascades, however it is very slow and takes about 11 seconds to detect the cars, so we are looking into better options.

We also need the object detection algorithm to complete the final trained optimization ML model, however we can still make progress on the basic ML code without that at the moment, so there are still no delays yet.

The simulation took longer than expected to finish, however it is finally done and we resolved the errors we were facing. We are afraid that the simulation might not run on our Pi due to it opening the GUI every time we run it, however we are looking into other Libsumo and other settings so we can run the sim without the GUI. Ankita and Kaitlyn are also going to try to get the sim running on the Pi early next week so we can resolve it earlier in case we run into any issues.

Changes to System Design

The main change to the design is that we no longer plan on using the Haar cascade we initially planned on using.

We also are no longer using both traffic APIs and instead plan on just using the TomTom API since we looked into it more and both provide similar data and TomTom updates more frequently.

Overall Takeaways and Progress

• New YOLO model working
• Simulation complete – with car spawning, traffic API inputs, and traffic light state changing functions
• The PCB has been ordered and is coming next week. We have also completed the Arduino code to control the LEDs and are in the process of testing the code.