MPPI Trajectory Visualization and Debugging

Source: ros2-copilot-skills MPPI trajectory visualization skill

Why This Matters

MPPI can feel opaque if you only look at the final cmd_vel and robot motion. Visualization is how you see whether the controller is exploring reasonable trajectory candidates, overreacting to costs, or being constrained by bad assumptions upstream.

Distilled Takeaways

• Visualizing sampled trajectories makes MPPI tuning dramatically more interpretable.
• Good debug views reveal whether the controller is considering useful candidates at all.
• Visualization helps separate critic problems from costmap or kinematic problems.
• The goal is not prettier RViz output. The goal is faster diagnosis of local-motion behavior.

Practical Guidance

• Use visualization when MPPI behaves strangely in only some environments or maneuvers.
• Compare what the controller samples with what the operator expected it to sample.
• Treat empty or obviously skewed candidate sets as a sign of deeper configuration issues.
• Pair visualization with repeatable test paths and scenarios.

Corroborating References

• Nav2 RViz and debugging docs
• Nav2 tuning guide

When to Read the Original Source

Go to the original skill when you want the concrete visualization hooks and the specific debugging patterns for understanding MPPI candidate trajectories in RViz.