nav planning

Product targets

Walking around SHZ office (premapped)

validates that:

• we can create map
• we can do loop closure
• we can load the map & relocalize & robot task in that frame
• it's not a small map :)

Walk around shenzhen SeaWorld

validates that:

• robust obstacle avoidance,
• a lot of dyanmic obstacles, people etc
• pretty large public space
• difficult to map

needs:

• better dog (mid360)

Multistorey nav

validates that:

• We can nav in 3D envs needs:
• better dog (mid360)

One robot maps other robot uses the map

needs:

• better dog (mid360)

Multi embodiment mapping

two robots mapping the same space, this space is somehow aligned globally (once robots see tags on each other, or once two maps relocalize against each other, etc)

Dynamic global 3D map

This week

• Go2 POC nav - dynamic 3d map, relocalization, real time PGO demo in the SHZ office

Jeff Mid360 dataset

• TODO mount on something flexible

• TODO mount on 45 degrees for Andrew

• TODO format a question for author of fastlio

• TODO let's get a fastlio author on a call

• IVAN REVIEW/Test on G1 Andrew Global + 2D A*

Tomorrow we pivot/sync

• TODO check sensitivy to process priority (try nice -20)

• TODO great multistorey / outdoors dataset on fastlio/mid360

• dumb raycaster version on G1

Andrew

• TODO needs dataset
• TODO wrap LoGlo + 3D Planner, maintaining only global graph, no global map (LoGlo is actually Terrainmap btw)

let's think about relocalization if no global map. (btw andrew graph builder can consume a random new large global map)

relocalizer would take LoGlo + Global map -> Relocalize

Ivan

• Real time PGO for go2
• Stream I/O stuff