Preparations for SPOT are almost coming to close. We ordered all our parts for the year and should be close to building and finishing our robot. Last semester we finished the chassis cad for SPOT and this semester we just have to focus on the launching mechanism and ball housing. Pictured below is the design we made for our ball housing. It operated very similarly to how a gumball machine works. Essentially a set of offset disks move the balls around inside and make it so that one ball gets loaded in at a time .Allowing SPOT to control how many balls it shoots at a time to its intended target. This loading mechanism will then be connected to a rail that feeds the balls into a flywheel launcher powered by two DC motors. The gumball machine loading mechanism will be powered by a single stepper motor that gets a control signal from the raspberry pi and powered by a separate LIPO battery since the Pi cannot output the required voltages for the motor. We will be using L298n Motor Drive Controller Boards to be controlling most of our motors since they allow for support of two dc motors per board or one stepper motor. In addition, we realized this semester we had to buy a portable battery pack for the raspberry pi in order for the robot to be mobile. We bought a typical usb-c power bank because it was a pretty cheap solution that could power the raspberry pi for a long time.
In terms of software for SPOT we are in the middle of training an XML file for it to recognize specific logos that we want it to target. However in the meanwhile, we have been able to do facial detection using Haar cascades and open source XML files trained to detect faces. We also experimented with the YOLO method to see if it would be a good fit for SPOT. Ultimately, we decided not to go with the YOLO method since it was simply too taxing for the Raspberry Pi and did not provide adequate results for real time objection detection. Using the Haar Cascade method we were able to get both facial detection and distancing working. For distance we just looked up the size of an average person’s face and depending on how big of a rectangle is made around a person’s face in OpenCV we are able to calculate the distance they are away from the camera. Our next goals are to work with the GPIO pins on the raspberry in order to be able to send a PWM signal to our motors, but until we have access to our robot we will have to put that on hold. Overall SPOT is coming up pretty well and we hope that it will be a success