Author: mwandles

This week, our team focused on finalizing our subsystems as well as starting integration. Not having the motor control board was a blow, but we still have our backup plan of direct connecting the battery to the launcher. Outside of directly connecting the battery to the launcher we experienced little to no design changes. During integration between the CV and the Motor Control, we found that we our robot wasn’t turning well either overrotating past the POI or rotating very slowly.  In order to counter this we decided to implement PID control. We also while testing found that running pose estimation when there were multiple people in the room caused a lot of latency. In order to counter this we implemented the temporary solution of capping the amount of people that we’d perform pose estimation on. We also realized that when actively targetting someone the latency that would come from running pose estimation on the target could interfere with our control, for that reason we decided to cut pose estimation once a target was found.

 

Our unit testing so far has been focused on making sure the CV is accurate by manually following our codes turning commands and seeing if our launcher is in line with our POI. The test we are currently running consists of the same except with the PID control fully controlling our launcher as the integration testing of our old linear turning algorithm failed.

This week, our team mostly worked on building our new structure. We drilled holes in our PVC sheet and attached the motors to it. Our robot also came in, so we worked on getting the motors attached to it as well with the PVC sheets. Our new motor board also arrived, so we have been working on integrating it. The most significant risk that could jeopardize the success of the project is still the launcher, but that will be revealed in the next few days as the new motor control board is integrated. We tested our launcher at 10% power and analyzed that it consistently reached around 1.5 feet. Assuming power and range are linear, this means that we can at least launch 15 feet. We are thinking of testing out different wheels with different frictions/compressions to see which one launches the tennis ball the best.

No changes were made to the existing design of the system from last week. There is also no change in schedule.

Over the new week we plan to refine our structure and get started on building the ball feeder/ramp as well. We also want to get the new motor board working and begin integrating it with everything else, like the CV and robot motors.

Here is our structure so far:

This week, our team spent most of our time preparing and doing the interim demos. We believe our biggest risk is still with our launch, as now we have to wait for our replacement motor board to arrive, but after our lower power tests this week we are more confident that it will launch our required distance of 20+ feet. This week we also have been considering a design change by the suggestion of Tamal. Since we have fallen slightly behind due to our parts breaking down, he made the suggestion to instead use a mobile robot for the turning of the base instead of our current turret design that we would have to build from scratch. This idea could both make implementing the turning easier and more reliable, but also opens up many more possibilities, such as actually moving towards a target who is out of range. We think this idea is promising and we have been researching what robot kit would be the best to use, so we can have it ordered hopefully by Monday. We would still likely use the PVC materials to place on top of the robot to give the launcher some height, but the actual lid would now be stationary with the robot doing the turning and moving instead.

Over the new week, once our new board and potentially our robot arrives, we want to focus much more on working with the motor control code and actually integrating all our subsystems now that we have tested them individually.

Over the next few weeks we will be continuing to run tests as we finish and integrate our subsystems. So far we have been able to test a low power version of the launch as well as the computer vision and physics calculations separately, all of which gave promising results. In the coming weeks as we integrate all these systems as well as implement the turning, we will also be running many tests on these. We will ensure that the range and physics calculations actually match up with the physical launch and tune them as necessary, and that our targeting actually properly tracks the target and our launcher faces them respectively. Based on the results of our tests, we will likely have to tune our targeting algorithms however much necessary to ensure the accuracy meets our requirements.

This week, the main thing we did as a team was place our first set of orders. We ordered all the parts necessary for the main launch mechanism , which we plan to have done by the interim demo. We already started building it this weekend and so far it is going smoothly. On top of that, we all individually worked on our sections of the project and the Ethics Assignment.

As usual, we still believe that the launcher mechanism is the biggest risk of our project, but we are doing our best to get that part done as soon as possible to mitigate this risk. If we are confident in this by/before the interim demo, we will have confidence that everything else can come together as well.

The only changes made to the existing design of the system is that the CV, depth AI, and target coordination will all be done on the Oak-D Pro camera’s processor rather than the Raspberry Pi, which is just as capable and there will be less back and forth between the components. However, there is no change in costs and the overall project design is still the same as before.