The reason we failed tonight:

(1) Space decoration:

• it took us 40 mins to setup everything (new floor lamp, super long power strip, table, blimp, chair and so on)

• It sitll not a good setup. For example, the power strip is at the center of the field, which is easily stumbled.

• The floor lamps were placed on the floor, the direct light was so bright that the light source is uneven.

(2) Blimp setup:

• We stacked the three heart shape balloons on top of the fish balloon, the overal height of the blimp is too high which the top heart shape balloon will easily get in touch with the net. In this case, the friction will slow down our blimp, which is not good enough.

• The heart shape balloons have a very serious leak problem, and we had to change the plasticine payload once a while.

• The latest green fish balloon will interfere the color detection of the OpenMV cam.

(3) Control:

• Color detection: since we never tested the green balloon color detection in the simulation room, we spent a lot of time tunning the color threshold.

• Green ball capture: it was really hard to capture the green ball even if we place the green ball in front of the blimp since there is some design offset between the basket position and green ball.

• OpenMV cam: OpenMV cam worked not very well at first. Sometimes it would just stop working. The wire connection and the without pull-up resisters are the main problems. After we added two pull-up resisters, things became much better.

• Vertical control: since the blimp itself (mainly heart-shape balloons) has the leak problem, it was hard for the previous vertical control to do the right thing.

• Lidar detection: sometimes the Lidar sensor at the nose of the blimp was not working. We think the wire connection has some problems. But it's hard to tell.

• Upload control code: Everytime we tried to changed control code, we had to connect the wire with the laptop to do it, which may accidentally mess up the wire connection.

• Debug problem: it was really hard for us to debug on-site since the upload control code process is really painful and time consuming.

• April tag Integration: after we manually put the green ball into the basket, the blimp could conduct seeking but it was hard for the camera to see the April tag since the green balloon can block the line of view. It has very little chance of seeing the April tag, even if it sees the April tag for several seconds, the rotation inertia will make it lost the target real quick.

Action items of improvement:

(1) Space decoration:

• Arrange the power strip wiring
• Put the floor lamp in right place(proper heigh), and the angle of the lamp should face to the ceiling. Diffuse reflection from the ceiling will provide sufficient brightness.
• Move a big table from the hallway into the simulation room, we really need that for placing our laptop and other equipment.

(2) Blimp setup:

• We need to build a stable blimp platform that everyone could test and debug (no serious leak problem in a short time, the overall dimension is not too high)

• Build more battery cases and put them on the back so that we could use the battery to balance the basket at the nose.

(3) Control:

• We need to setup a communication (peer to peer or mesh network)between the laptop and esp32. In this way, we could easily debug the code.

• We need to have a led to see the detection stage

• We need to make the IMU and barometer work

• We need to solder the camera output wires so that they won't easily get loose.

• We need to put the basket on the right place so that after it capture the ball it won't block the camera to detect the April Tag.

Next Post Previous Post