EF Plane SLAM Back-end Benchmark

This repository contains all required scripts and notebooks to compare quality of different SLAM Backends using planes as landmarks.

Data

As the data source we provide two variants: EVOPS dataset and synthetic planar poses generator from mrob library.

Supported backends

For comparison needs two algorithms except origin EF backend are implemented in mrob library. They are:

• Pi-Factor
• Bareg
• EF Eigen Factors

Evaluations

You can see examples of backend comparisons in two Python notebooks:

• eval_icl.ipynb --- example of backend comparison on data from EVOPS dataset (sequence based on ICL NUIM Living Room kt0 trajectory)
• eval_synthetic.ipynb --- example of backend comparison on synthetic data generated by CreatePoints method from mrob library

Metrics

To compare SLAM backends predicted trajectories on perturbed origin poses are compared with ground truth poses using classic ape and rpe metrics.