The pre-planned touchdown point comes from orbital imagery whose resolution can hide boulders, crater rims and steep slopes below the lander's own safety margins. In naive mode the guidance flies straight to that fixed point no matter what is actually there. In TRN mode, once the lander crosses the hazard-detection altitude, a simulated downward-looking sensor sweeps the ground within its scan radius, every boulder/crater/slope inside that circle is added to a live hazard map, and the guidance searches nearby clear ground for the closest safe patch — then bends the remaining trajectory toward it.
- Hazard density — how cluttered the unmapped terrain is; denser fields make the naive target more likely to sit on or near a hazard.
- Sensor scan radius — how far the onboard sensor can see from the lander's position at the detection altitude; a wider radius finds safer, closer alternatives.
- Descent speed — playback pacing only, for comparing runs quickly.
Real-world relevance: this is the essence of Terrain-Relative Navigation / hazard-relative navigation as flown on Mars 2020 (Perseverance) and demonstrated by NASA's ALHAT program — because Earth-to-lander light delay rules out real-time human piloting, the retargeting decision has to be made autonomously, onboard, in the last tens of seconds of descent.