Drone
Constraint Update
- We had to go with a bigger deisgn
- Limited flight time
Ordering
- We have an order from eBay
- Parts list (and the parts document)
Frame customization
- Frame is a bit heavy (680 g) but there are certain parts that we can replace
- We will CAD the mounts for the CP.
Zedboard / ML
FreeTPU
FreeTPU is a hardware accelerator for machine learning that fits on to the ZedBoard FPGA. Comes with API, we can look into customizing it for our purposes. For YOLO V3 has a inference time of 0.2 s (in theory frame-to-frame time).
Justifcation design:
- Why FreeTPU?
Risk:
- Still uncertain about the overhead between opertions.
- Compilation on FreeTPU presents constraints on where we can compile. The API should be compiled on a ARM chip device (Zedboard or Pi).
Flight Mode
Risks:
- currently both the PLB and PMB boots linux but require after-boot commands, we need to figure out how to execute them automatically upon power on
- an indication that the computing sytem is good to go. As a constrain, there is no way to tell the FC that the CP is flight-ready.
Base Station GUI program
Arthur made a prototype in C#.
Risk: there is a fragmentation of code base that could restrict teammates’ ability to be agile.
Proposed mitigation: a standard protocol for program interfaces. As long as we have a solid protocol, any team member have the freedom to experiment with whicherver language they like.
Risks Update
| Risk | Likelihood | Severity | Risk Index | |
|---|---|---|---|---|
| Drone flight hardware (flight controllers, radio, motors) cannot function due to crashes and/or damage. | 0.9 | 1.0 | 0.9 | |
| Payload is too heavy which significantly increases drone motor requirements and significant reduction in flight duration. | 0.9 | 0.6 | 0.54 (v) | |
| Accidents that damage the drone and computation equipment that require extra budget that we may not have. | 0.6 | 0.9 | 0.54 | |
| Total loss of drone hardware and payload during flight. | 0.5 | 1.0 | 0.50 | |
| Not enough time commitment from team members. | 0.7 | 0.7 | 0.49 | |
| Other courses and obligations will take too much time away from capstone progress. | 0.8 | 0.6 | 0.48 (^) | |
| Legacy documents for the project are insufficient, resulting in poor maintainability/extensibility for the client. | 0.8 | 0.6 | 0.48 | |
| Payload is too heavy which exceeds total take-off weight. | 0.4 | 1.0 | 0.40 | |
| Financial inefficiencies leading to budget overruns or lack of capital. | 0.6 | 0.5 | 0.30 | |
| Underestimation of project scope or work required, leading to insufficient time management and burn-outs. | 0.5 | 0.6 | 0.30 | |
| Constrained to purchase lower-quality components due to budget, resulting in lower performance. | 0.6 | 0.5 | 0.30 | |
| Team is indecisive or cannot make a timely decision — resulting in delay. | 0.4 | 0.6 | 0.30 | |
| Team lacks ineffective communication skills which lead to overlapping work, missed work, and/or incompatible work. | 0.7 | 0.4 | 0.28 | |
| Development and management technique/methodology is not effective, leading to productivity losses. | 0.4 | 0.7 | 0.28 | |
| Not enough time to work on documentation. | 0.7 | 0.4 | 0.28 | |
| Footage from the camera is not stable or clear enought for image processing. | 0.6 | 0.4 | 0.24 (^) | |
| Failure to acquire regulatory compliance resulting in inability fly drone legally. | 0.3 | 0.5 | 0.21 | |
| The software, tools or development environment for the project is inadequate. | 0.4 | 0.5 | 0.20 | |
| Knowledge and skill regarding ML is insufficient. | 0.5 | 0.4 | 0.20 | |
| Technical debt paydown impacts project timeline. | 0.4 | 0.5 | 0.20 | |
| Not enough test cases to validate design. | 0.5 | 0.4 | 0.20 (^) | |
| Deliverables fail to meet client’s expectations. | 0.2 | 0.9 | 0.18 | |
| Access to tools and shops for modifying and repairing drone hardware is inadequate or non-existent. | 0.2 | 0.8 | 0.16 (v) | |
| Find a venue / large indoor spaces to test fly the drone without legal actions. | 0.5 | 0.3 | 0.15 | |
| Internal documentation or documentation for libraries and parts are not sufficient for development. | 0.3 | 0.4 | 0.12 (v) | |
| New technology or research emerges, changing the scope significantly. | 0.2 | 0.5 | 0.10 | |
| Sabotage of the project. | 0.01 | 1 | 0.01 | |
| Sudden loss of client. | $\leq$0.1 | 1 | 0.10 | |
| Client demands modification to the scope and requirements of the project that leads to delays or feature cuts. | 0.1 | 0.9 | 0.09 (V) | |
| Sudden loss of team member. | $\leq$0.1 | 0.9 | 0.09 | |
| FPGA board lacks documentation. | $\leq$0.1 | 0.8 | 0.08 | |
| Client is not cooperative or does not provide necessary information. | $\leq$0.1 | 0.8 | 0.08 | |
| Key components are not available. | 0.1 | 0.7 | 0.07 | |
| Purchased orders of equipment or tools delayed or lost. | 0.1 | 0.5 | 0.05 | |
| Client is not available enough to provide significant help. | $\leq$0.1 | 0.5 | 0.05 | |
| Lack of resources to acquire machine learning knowledge. | $\leq$0.1 | 0.5 | 0.05 | |
| Camera module fails to interface with FPGA. | $\leq$0.1 | 0.5 | 0.05 | |
| Data transmitter fails to interface with FPGA. | $\leq$0.1 | 0.5 | 0.05 | |
| Market competition significantly affects project requirements and scope. | $\leq$0.1 | 0.4 | 0.04 | |
| Software license does not allow our application to be delivered. | 0.1 | 0.3 | 0.03 | |
| Laws regarding drone operation and piloting change significantly. | 0.1 | 0.2 | 0.02 | |
| Not enough FPGA logic elements to implement a desired ML model. | 0.5 | 0.0 | 0.00 (V) | |
| Not enough machine learning training data. | 0.5 | 0.0 | 0.00 (V) | |
| Camera module lacking in documentation. | 0.0 | 0.8 | 0.00 (V) |
CAD
