Baseline Performance

Overview

This document describes the expected baseline performance of MOLA SLAM on unperturbed data.

Baseline ATE (Unperturbed Data)

Baseline ATE: ~23cm (0.23 meters)

This is MOLA's natural localization error when processing clean, unperturbed LiDAR data from the Isaac Sim simulation.

Baseline Performance

MOLA achieves ~23cm ATE on our test trajectory, which represents good SLAM performance:

For comparison, typical MOLA performance: - Ideal conditions (real robot, optimized params): 10-20cm ATE - Good conditions (our case): 20-30cm ATE - Moderate conditions: 30-50cm ATE - Poor conditions (degraded sensors, fast motion): 50cm+ ATE

Baseline Characteristics

Running MOLA on unperturbed data: - Frames processed: 113 frames (10Hz LiDAR) - Keyframes selected: 49 keyframes (MOLA selects when robot moves) - Trajectory length: ~11.3 seconds - Points per frame: ~50,000 points - Loop closures: 2-3 successful loop closures

The baseline provides a reference point for measuring adversarial attack effectiveness.