Announcing the first production robot navigation framework on $500 hardware Explore the world once → your robot agent will relocalize and build a persistant, spatial memory across sessions SLAM, relocalization, loop closure, map i/o, planning, control No ROS. Open source.

Dimensional v0.0.13 has dropped. Get building 🦾🔥

2/ Robust navigation is a key requirement to deploy humanoids and quadrupeds in production Current nav frameworks work in the lab but fail as you scale to large maps, multi-day sessions, multi-story deployments…due to map drift, 2d only, static/not updating, requires premap

3/ Full stack robotics companies today build navigation from scratch for their custom hardware. Local/global planning, trajectory control, recovery behaviors, SLAM, much more Developers on dimOS can now strap a cheap lidar to ANY robot & save 6 months of nav engineering ⬇️: loop closure & relocalization tuning with apriltags

4/ We extend loop closure (from above) to also work on geometric and semantic features at any dynamic set of intervals. runs both synchronous and async depending on compute constraints. each line is a loop closure event and shows the delta in old path → fixed path

@dimensionalos no ROS you say? I already love it

@dimensionalos pip only ;)

5/ With multi-story planning and trajectory control supported across all embodiments More tooling coming soon to allow agents/autoresearch to fine tune trajectory config hyperparameters on a replay so developers don't have to tune manually / randomly guess on values

@dimensionalos DimOS is really great to work with, especially as someone who is from an adjacent field getting started in robotics / physical AI. Really looking forward to this release

@dimensionalos 🔥

@dimensionalos 🤠

@dimensionalos Congrats!!! This is awesome

@dimensionalos 1 what's the license? 2 can I integrate my post quantum cad? 3 need naturalistic movement controls

@dimensionalos apache 2.0

@dimensionalos persistent spatial memory across sessions on $500 hardware — we're solving the same problem for CAD agents with CoMeT. the robots navigating your map still had to be designed first:

@dimensionalos have i been here before..

@chris_j_paxton @dimensionalos woah 👀

@dimensionalos Pretty neat! Is $500 reference to the dog or compute (orin?)

@dimensionalos ~$200-500 for the lidar, $1000 for the quadruped. Compute negligible as no gpu requirement.

@dimensionalos Okay @maticrobots so we'll see this in Stable 166?

@dimensionalos Full SLAM and navigation on $500 hardware with no ROS is honestly a great price point

@dimensionalos This is sick!!!

@dimensionalos cross-session memory is the right hard problem, but relocalization isn't where it breaks. staleness is: the map's from last session, a chair moved, and the robot trusts a world that's gone. catching stale memory beats storing it.

@dimensionalos yeah using relocalization to solve for / publish accurate transform frames is what we use to fix semantic object map staleness

@dimensionalos no ros is the flex tbh

@dimensionalos we see this as direct input for our skill-routing layer. The persistent spatial memory becomes a portable policy anchor across embodiments.

@dimensionalos

@dimensionalos Persistent spatial memory is cool. The next challenge is giving robots confidence in the environments, services, and agents they interact with.

@dimensionalos github?

@dimensionalos Is lidar necessary Does is support vio

@dimensionalos stereo nav is coming. don't support currently.

@dimensionalos Sooo dope

@dimensionalos $500 for SLAM with no ROS? What hardware specifically?

@dimensionalos What sensors it uses?