Issue 1: Hardware
| Item Number | HW 1 |
| Date Initiated | 28 March 2023 |
| Date Resolved | 05 April 2023 |
| Participants | Sankalp @schopkar, Sudhansh @syelishe |
| Description | Castor wheels of Khepera IV falling off during testing |
| Options | 1. Add foam padding to dampen the buckling effect 2. Decrease performance requirement on gap size from 21 cm to 19 cm 3. Re-print enclosure with stoppers to reduce vertical displacement |
| Resolution | Short term we went with (1), long term we will reduce the requirement to preserve hardware since it is borrowed from the AAR Lab |
| Justification | Redesign this late would take significant time, and there is no guarantee that the other solutions will work well without testing |
| Item Number | HW 2 |
| Date Initiated | 04 April 2023 |
| Date Resolved | 05 April 2023 |
| Participants | Sankalp @schopkar, Sudhansh @syelishe |
| Description | 3D printed enclosure damaged due to excessive heating of solenoid lock |
| Options | 1. Replace the solenoid lock 2. Change the control logic to decrease the current drawing time for the solenoid lock |
| Resolution | We proceeded with option (2) |
| Justification | Changing a component will require a significant design change |
Issue 2: Software
| Item Number | SW 1 |
| Date Initiated | 16 February 2023 |
| Date Resolved | 27 February 2023 |
| Participants | Harishankar @hravisan |
| Description | The controller is too slow and erratic in appearance |
| Options | 1. Improve the PID tuning for each action taken by the controller (moving or turning) 2. Decrease the density of waypoints generated by the planner 3. Move to an MPC controller which does not require close waypoints for smooth trajectories |
| Resolution | To remain on track, we are attacking on two fronts – tuning the PID while working on an MPC implementation in parallel and go with whatever works best at the end |
| Justification | Ideally, MPC would have been a good choice but we didn’t have the knowledge at the time to implement it, so we went with PID, in the beginning, to get a working baseline to iterate over. |
| Item Number | SW 2 |
| Date Initiated | 28 March 2023 |
| Date Resolved | 05 April 2023 |
| Participants | Dhanvi @dsreeniv, Sudhansh @syelishe |
| Description | Multi Label A* versus Sequential A* |
| Options | 1. Use sequential A* for each goal in an agent’s queue with CBS running for each new location popped 2. Use Multi-Label A* to determine a global path through each goal location and run CBS over the full path |
| Resolution | We went with (1) due to simplicity of extending and debugging. This moves away from optimality but is faster and in practice works well enough on hardware |
| Justification | To save time and effort for resolving more complicated edge cases and ensuring planner doesn’t break |
