Scale Invariant Multi-agent Map Merging

Method for merging graphs from multiple agents where the scales between them are unknown.

Dependencies

• Modified version of GTSAM with python bindings (hosted at https://github.mit.edu/vnav-project/gtsam)
• numpy
• numpy-quaternion
• sophuspy
• scipy
• numba
• tqdm
• fmc in find_max_clique/fmc

Setup

Update your PYTHONPATH environment variable to include the root of the project

export PYTHONPATH=${PYTHONPATH}:$ROOT where $ROOT is the full path to the root of this repo.

---

Usages

Create a multi-agent g2o file from a single-agent g2o file

1. cd process_g2o
2. python3 convert_to_multi.py ../datasets/manhattanOlson3500.g2o output.g2o

This will create a multi-agent g2o called output.g2o from the manhattanOlson3500.g2o dataset.

---

Perform pose graph optimization on a multi-agent graph

python3 multi_agent_optimization.py process_g2o/output.g2o output.g2o

This will perform the pose graph optimization on the multi-agent graph in process_g2o/output.g2o, and output as output.g2o in current directory.

---

Build Adjacency

python3 build_adjacency.py datasets/city10000.g2o

where input.g2o is a multi-agent g2o file, and output.g2o will be the optimized one.

---

Style Guidelines

• python: https://github.com/google/styleguide/blob/gh-pages/pyguide.md
• C++: https://google.github.io/styleguide/cppguide.html