Meeting Minutes - Path Optimization
October 27th, 2022
The whole task should be completed by around January
Went over CONOPS document (https://www.aerialevolution.ca/wp-content/uploads/2022/10/2023-AEAC-Competition-RFP-v2.0-17-Oct-22.pdf)
Important points: 8, 9, 14 - 26
Â
Task 1: Long distance flights, with a rough path given at start of flight. During flight, the ground team will receive data on Waypoints to avoid, and waypoints the drone needs to reach.
Example path: Follow route Waypoint 3; Waypoint 14; Waypoint 6; Waypoint 12; Waypoint 1
Sample Message: Avoid the area bounded by Waypoint 3; Waypoint 7; Waypoint 4; Waypoint 2. Rejoin the route at Waypoint 13
Brainstormed ideas: Dijkstra's algorithm
Potential issues: Speak with leads/other sub-teams on data communication (i.e. understanding how we will get the data from object detection)
Â
Task 2: On-Demand Passenger Transport. Routes will be given beforehand. We want to use an algorithm to compute the path beforehand (bonus points).
Sample Message: Route number 1; 2 pers; Waypoint 12; Waypoint 3; 15 kg; obstacle 2 m to NE; $112 Route number 2; 6 pers; Waypoint 3; Waypoint 7; 5 kg; nil; $50 Route number 3; 4 pers; Waypoint 7; Waypoint 8; 15 kg; other comment; $150
Brainstormed ideas: Dollar per unit time, spam combinations since infinite computation power
Overall approach: Since the drone must travel within 50m deviation of the straight, direct route between waypoints, we can represent each waypoint as nodes. Apply graph theory.
We also need to consider and experiment how the weights and number of people will affect the drone as well as understand the effect of wind (can be added to the algorithm later on)
Â
TODOs
Do research on potential algorithms and look for research papers, so we can finalize on a plan of action in the coming week.
Speak with other sub-teams.
Think about whether we will need a Reinforcement Learning model or can we work everything out with path optimization algorithms.