Global Path Planning

The Global Path Planning module computes the optimal racing line for the complete track after the first lap is finished. It processes the full track map from SLAM and generates a single time-optimal trajectory that is used for all subsequent laps.

Overview

After the vehicle completes its first exploration lap, SLAM publishes a complete map of the track. The global path planner uses this full track knowledge to compute the fastest possible racing line, considering the entire lap topology and optimizing corner entry/exit speeds.

Key Features:

• Full-track optimization: Plans 100 waypoints covering the entire lap

• Time-optimal racing line: Minimizes lap time with complete track knowledge

• One-time computation: Runs once when global map is available, then shuts down

• Closed-loop path: Generates repeatable trajectory for all remaining laps

• 10-15% faster: Significant improvement over local planner paths

Operational Phase:

• Inactive: During first lap (local planner active)

• Active: After global map received (computes optimal path)

• Shutdown: After publishing path (job complete)

System Flow

SLAM Global Map (after lap 1 complete)
     ↓
Cone Ordering (sort all cones into closed loop)
     ↓
Trajectory Optimization (time-optimal over full lap)
     ↓
Path Interpolation (convert to constant 100Hz grid)
     ↓
Publish Path + Completion Signal
     ↓
Node Shutdown (no longer needed)

Comparison with Local Planner

Aspect	Local Planner	Global Planner
Horizon	10 waypoints (~10-20m)	100 waypoints (full lap)
When Active	First lap only	After first lap
Solve Time	60-180ms	1-5 seconds
Update Frequency	Every SLAM update (~1-5 Hz)	Once per race
Path Type	Open-ended short segment	Closed-loop full lap
Performance	~10-15% slower than optimal	Optimal racing line

Global Path Documentation

• Overview
  • Purpose
  • Comparison with Local Planner
  • System Architecture
  • Key Characteristics
  • Track Representation
  • Integration with SLAM
  • Optimization Time
  • Output Format
  • Typical Performance
  • Robustness
  • Node Lifecycle
  • Transition from Local to Global
  • Limitations
• Configuration
  • Settings Structure
  • Key Differences from Local Planner
  • Optimization Parameters
  • Constraint Limits
  • Solver Configuration
  • ROS Topics
  • Example Configurations
  • Performance Tuning
  • Debugging
  • Comparison with MPC Settings

Published Topics

Topic	Type	Description
/path/global	feb_msgs/FebPath	Optimized global trajectory (full lap)
/path/finished	std_msgs/Bool	Signals global path is ready (True)
/path/global/viz	sensor_msgs/PointCloud	Path waypoints for visualization

Subscribed Topics

Topic	Type	Description
/slam/map/global	feb_msgs/Map	Complete track map (triggers optimization)
/slam/state	feb_msgs/State	Current vehicle state (for initialization)

Performance Characteristics

Computation Time:

• Cone ordering: 10-50ms

• Trajectory optimization: 1-5 seconds

• Path interpolation: 10-20ms

• Total: 1-5 seconds (one-time)

Path Quality:

• True optimal racing line (within solver tolerance)

• Friction-limited cornering speeds

• Optimal braking/acceleration points

• Lap time: 10-15% faster than local planner

Problem Size:

• Decision variables: 700 (7 per waypoint × 100 waypoints)

• Constraints: 1000-2000 (dynamics, bounds, optional cone avoidance)

• Solver iterations: 50-200 (depends on track complexity)

Optimization Approach

The global planner uses the same time-optimal formulation as the local planner:

Objective:

\[\min \sum_{i=1}^{100} \Delta t_i\]

Subject to:

• Bicycle dynamics constraints

• Track boundary constraints (closed loop)

• Velocity, acceleration, steering limits

• Friction circle constraint

Key difference: The optimization considers the entire lap as one connected path, allowing it to plan optimal corner entry/exit speeds by “looking ahead” to future sections of the track.

Node Lifecycle

The global planner has a unique lifecycle designed to save resources:

1. Initialization: Node starts and waits for global map

2. Triggered: Receives /slam/map/global message

3. Optimization: Solves full-lap trajectory (1-5 seconds)

4. Publishing: Outputs global path and completion signal

5. Shutdown: Calls self.destroy_node() and exits

This design ensures the global planner only consumes resources when needed.

Integration with System

Transition from Local to Global:

1. Vehicle completes first lap

2. SLAM publishes /slam/map/global

3. Local planner receives message and shuts down

4. Global planner receives message and starts optimization

5. Vehicle continues tracking previous local path

6. Global planner publishes new path

7. MPC seamlessly switches to tracking global path

8. Global planner shuts down

MPC Integration:

The MPC controller subscribes to both /path/local and /path/global. It tracks whichever path is most recently published, making the transition seamless.

API Reference