Further Work Section

[PeterDeJ17]

Kind of done needs review - struggled to think of many more things that could be done?

[kmiller237]

Everyone should read and add to this

[kmiller237]

Single-agent maze solving:

• Localise within map
  • EKF
  • AMCL?
• Path planning
  • Probably A* algorithm, modified to allow for width of robot
• Exploration
  • Start by scanning environment/periodically scan to improve map?
  • Greatest unknown frontier as next destination for path planning algorithm?
  • Depth first to avoid doubling back?
• Behaviour once goal detected
  • Return to starting point or continue searching for more goals?

Co-operative maze solving:

• Make comms system more flexible
  • Decode messages into arbitrary ROS message types (e.g. specify that message for synchronising sonar is of type float)
  • Allow responding to certain message types without line of sight
• Synchronise sonar scans
  • Hard-coded bands initially
  • Dynamically allocate bands
• SLAMOT (SLAM with Object Tracking)
  • Improve sonar detection of robots (e.g. by covering their bums with a bit of plastic)
  • Avoid mapping other robots
  • Track positions of other robots
• Coordinate movements
  • Simplest strategy: explore away from other robots
  • Announce goal found
  • Request assistance?
• Joint mapping (loop closure)? (edited)

[kmiller237]

As far as I can see its "Synchonise sonar scans" and "Make comms system more flexible" that could do with more detail just to show we have thought about it? Maybe just a couple of sentences on each?

[kmiller237]

Andrew did this so think it can be closed?

[AMFagan]

I left it open because it's such an important section, but if most of the team have now reviewed it then sure