Our project has received a significant shakeup over the last week. Last Sunday, we put a significant amount of additional work into our proposal presentation compared to when we had finished before we wrote the last report. This resulted in a shakeup of both our schedule and roles for the 3 of us. We consider the proposal presentation to be of great value, as it forced us to think a lot more about what we are doing. Our new schedule for the individual portions of the project is in the project proposal gantt chart.
In addition, the inventory being shared on Wednesday changed some of our part and depth-sensing plans. Two depth-sensing cameras are available, which may reduce our need for additional distance sensing hardware.
Risk 1: Hardware component availability and integration
We’re currently evaluating robotics chassis options to ensure adequate payload capacity, 2-3kg and platform size for the KR260. We are mitigating this by researching multiple vendors and chassis types.
Risk 2: Model training timeline and FPGA resource constraints
Training YOLOv3-tiny and fitting quantized weights on FPGA could take longer than anticipated. Mitigate this risk by starting model training this week ahead of schedule. The back up plan is to use pretrained weights if custom training fails. fall back to Vitis AI if custom RTL integration proves too difficult.
Risk 3: CPU processing constraints
We are requiring the CPU on the KR260 to convert frames into 1D arrays, handle robotics control logic, and perform postprocessing of frames while evaluating depth. We will mitigate this risk by starting with simply bounding boxes for our distance measurements. If we have the processing overhead, we would then use a depth camera system for more accurate distance measurements. In the absolute worst case, we would need to offload some of the CPU responsibilities to a separate board, which would introduce difficulties in making the two boards coordinate.
No photos yet as we’re in the planning phase, but hardware selection will be finalized by next week.
