[Feature]: Improvements in local planner/ Navigation

[prat1kbhujbal]

Problem

1. Lots of values are hardcoded: eg: speed, tolerances, etc
2. If robot is too close to obstacle it gets stuck, maybe add more planner states like recovery methods or improve the obstacle avoidance at least for dynamic obstacle
3. Restructure navigation configs to take from single config file, easier to modify and more readable
4. I cant simulate but not sure if it happens on real robots with sudden stop or start of the robot for jerky motion as there is no acceleration or velocity profile
5. Add more debug logs for why robot abandon its current goal
6. Improve path smoothing currently robot moves in zig-zag way
7. Add more local planners like teb, dwa ?

8. dimos/dimos/mapping/occupancy/gradient.py

   Line 53 in a035fb3

   distance_cells = ndimage.distance_transform_edt(1 - obstacle_map)

   
   These lines calculates over full grid, this will cause more ram/cpu usage

Proposed Solution

Solution:

• Add config file
• Add velocity smoothing or velocity profile depending on obstacles, distance to goal and generated paths
• Log reason codes whenever goals are abandoned or replanned
• Add better path smoothing
• Integrate other local planners like dwa or teb for flexible motion planning
• Use local window around the robot to update map

Alternatives Considered

No response

Additional Context

No response

---

Synced from DIM-704

[leshy]

Hi, which robot? which blueprint?

[prat1kbhujbal]

I tested with unitree go2 in simulation