Meeting 5

This is part of a series of posts that summarize my meetings with Professor Ian M. Mitchell for our own records.

During the fifth meeting, on March 9th, we discussed the following.

This week’s highlights

• Refined drafts of the following based on previous discussions:
  • State space & boat model
  • Evaluation scenario
• Composed the following drafts:
  • Planning architecture overview
  • Perception (early draft)
• Made progress towards using K. Wille’s Masters thesis materials as a platform for controller & micro/local planner evaluation.
  • Extending it to support the evaluation scenario.
  • Work on an initial, naive micro planner implementation.
• Extensive progress in developing a planet-level planner.
  • Pathfinding between any two points on the planet works and is highly performant.
  • Highly extensible, modular design created with testability in mind.
  • Added a CLI.
    • Has a .kml output option for easy display on Google Maps.
  • Project will also be used for the production-intent realtime planner.
    • There are many general-purpose utilities required, maintaining a common library with a mono-repo helps ensure that nothing diverges.
    • This will make E2E testing easier and maintaining cross-system compatibility easy.
      • Unit tests that use both libraries can be easily implemented.
      • A target containing all planners would simple to make.
    • The global planner can be used to bootstrap micro/local.
      • The current A* implementation will be useful for local.
• Further readings from Siegwart’s Intro to Autonomous Mobile Robots, specifically relating to sensors.

Summary of discussion

• Reviewed the above.
• Decided to shift focus to developing a Global solution first in order to ensure that one of the system components is completed as a deliverable by the end of the term.
  • Postpone further development of the micro/local planner (on vessel).
• Discussed:
  • Ada 1.0’s global path planner
  • The global planning problem
  • The current geodesic grid approach to global
  • Alternative methods (e.g. Fast marching, Fast sweeping)

Goals for next meeting

• Define the global pathfinding problem (mathematically).
• Read the following:
  • Ian’s IROS 2008 Workshop Slides
  • Static / time-independent HJ PDE Fast Marching Method
  • Ken Alton’s thesis (first two chapters)
  • Chapter 2.7 of the Toolbox of Level Set Method manual
• Further development of the global (planet-level) planner (on server).
  • Augment the geodesic graph by making distant neighbours available (neighbour reach $n$). This will double (or better with $n > 2$) the angular resolution available to the vessel at any time, offering the following benefits:
    • Significantly reduce the tendency for the path to travel along the geodesic grid’s original geometry.
    • Increase the accuracy/benefit of modeling vessel performance against apparent wind.
    • Note: watch for problems with hitting obstacles by “hopping” over them.
  • Display generated paths on the Viewer (OpenGL GUI application).
    • Use an additional GL_ARRAY_BUFFER containing a mapping of colour to each edge.