16 Aug

When we use POST method to create data to send it to the server, we do not need to list the models using known_models.py. Previously, we needed to put the name of the component in known_models python script like this: Then, we need to run the thumbnails python script to create the thumbnails...

15 Aug

In the past 2 weeks, I tried to develop and algorithm that can be used if we have a blimp that have a camera facing down. With the camera, we would be able to detect the field of interest and control the robots within the field. In this case, the current algorithm can be used to control any vehicle...

15 Aug

One thing that I wanted to do this week was to validate the effect of including frequency diversity in the network that we made for our robots. So, I programmed one ESP32 feather board with a 433 MHz packet radio on its stack to send out a lot of garbage values over the 433 MHz band extremely quickl...

14 Aug

During SURP 2021’s week 8, Shahrul Kamil bin Hassan (Team Arnhold), Bhavik Joshi (Team Arnhold), Marisa Duran (Team Arnhold), and I integrated the three epics in the Arnhold project which are Mesh Networking, Vision Processing, and Feedback Controlled Cars. The integration can be seen below. A...

14 Aug

During SURP 2021’s week 8, Bhavik Joshi (Team Arnhold), Marisa Duran (Team Arnhold), and I managed to integrate the vision processing and feedback controlled cars epic. Using Marisa’s OpenMV code and my code for controlling the cars, we were able to put an OpenMV camera in one of my four-wheeled o...

14 Aug

During SURP 2021’s Week 7, Bhavik Joshi (Team Arnhold) and I were able to create a swarm of car robots using his Mesh Networking code and my cars. The cars were controlled by a single joystick through Sudarshian Seshadri’s (Team Arnhold) React-app dash. A gif of the car swarm is attached below. ...

11 Aug

The video in the last post depicted a very slow detection of the ball. I also tested a CPU-trained model but, as expected, it was quite slow: To improve the performance of the algorithm, I replaced the original FasterRCNN model (torchvision.models.detection.fasterrcnn_resnet50_fpn()) with a mo...

11 Aug

Yesterday, I successfully integrated the camera with the dashboard (and the ESP32 featherboard). In the video demo below, I have the featherboard calling functions on the openMV camera, asking it to detect an apriltag and return rotation information (x, y, and z euler angles from the apriltags). The...

11 Aug

I merged the ball detection code with the distance estimation code and added a camera component to make it live. The following video shows the code running live on CUDA: Model prediction is very slow on the CPU; I am currently checking out techniques such as quantization to optimize the model. A...

09 Aug

We simulate the BOEM algorithm against two other offline algorithms. Since these two algorithms requires a long time, we simulated them with sliding windows. In the first scenario, the block size of BOEM started with 9 sec, and grows geometrically. The sliding windows for both EM and optimizati...