Topic 2: Planning & Navigation Systems

Topic 2 introduces the navigation stack (maps, localization, planners, controllers) and shows how to build waypoint-based navigation and room traversal for your humanoid.

2.1 Navigation Stack Overview (Module A)

A typical ROS 2 navigation stack (e.g., Nav2) consists of:

• Map:
  • 2D occupancy grid or 3D map from SLAM (Chapter 4).
  • Marks free, occupied, and unknown areas.
• Localization:
  • Estimates the robot’s pose on the map (e.g., AMCL, VSLAM output).
• Global planner:
  • Computes a path from the current pose to a goal pose across the map.
• Local planner / controller:
  • Follows the global path while reacting to local obstacles and dynamics.

This stack fits directly into your agent architecture:

• High-level tasks specify goals (e.g., “go to kitchen” → map coordinates).
• The navigation stack turns those into collision-free trajectories.
• Controllers execute the trajectories in simulation and on hardware.

2.2 Global vs Local Planning (Module B)

Global Planning

The global planner:

• Uses the full map to compute an overall route.
• Optimizes for criteria like path length, safety margins, or energy.
• Produces a static or slowly-changing path from start to goal.

Characteristics:

• Planning horizon: long (room-to-room, across building).
• Needs a reasonably consistent map.
• Updated when goals change or environment changes significantly.

Local Planning

The local planner/controller:

• Looks at the robot’s immediate neighborhood:
  • Local costmap from sensors (LiDAR, depth).
  • Dynamic obstacles and short-term changes.
• Adjusts velocity commands to:
  • Avoid collisions.
  • Respect kinematic and dynamic constraints.

Characteristics:

• Planning horizon: short (a few meters or seconds ahead).
• Runs at higher frequency than the global planner.
• Performs dynamic re-planning as new sensor data arrives.

Together:

• Global planner answers “how do I get there?”
• Local planner answers “how do I move safely right now?”

2.3 Waypoint Missions (Module C)

With mapping, localization, and planning in place, you can define waypoint missions:

• Sequences of intermediate goals:
  • Room A → hallway → Room B → docking station
• Represented as:
  • Lists of poses in the map frame.
  • ROS 2 actions (e.g., repeated NavigateToPose goals).

Example missions:

• Room-to-room traversal:
  • Start in lab, traverse corridor, reach office.
• Patrol routes:
  • Visit multiple checkpoints, loop back to base.

You will:

• Script missions using ROS 2 actions and launch files.
• Test them in your digital twin (Chapter 3) with SLAM maps (Chapter 4).
• Feed success/failure status into higher-level task graphs (Topic 3).

By the end of Topic 2, you should be comfortable:

• Configuring a navigation stack on top of your SLAM maps.
• Distinguishing roles of global and local planners.
• Creating waypoint-based missions that allow your humanoid to move autonomously between locations.