trendingtechnology / taichislam Goto Github PK

View Code? Open in Web Editor NEW

This project forked from xuhao1/taichislam

0.0 1.0 0.0 4.5 MB

A 3D Dense mapping backend library of SLAM based on taichi-Lang designed for the aerial swarm.

License: GNU Lesser General Public License v2.1

Python 100.00%

taichislam's Introduction

TaichiSLAM

This project is a 3D Dense mapping backend library of SLAM based Taichi-Lang, designed for the aerial swarm.

Intro

Taichi is an efficient domain-specific language (DSL) designed for computer graphics (CG), which can be adopted for high-performance computing on mobile devices. Thanks to the connection between CG and robotics, we can adopt this powerful tool to accelerate the development of robotics algorithms.

In this project, I am trying to take advantages of Taichi, including parallel optimization, sparse computing, advanced data structures and CUDA acceleration. The original purpose of this project is to reproduce dense mapping papers, including Octomap, Voxblox, Voxgraph etc.

Note: This project is only backend of 3d dense mapping. For full SLAM features including real-time state estimation, pose graph optimization, depth generation, please take a look on VINS and my fisheye fork of VINS.

Demos

Octomap/Occupy map at different accuacy:

Truncated signed distance function (TSDF): Surface reconstruct by TSDF (not refined) Occupy map and slice of original TSDF

Usage

Install taichi via pip

pip install taichi

Download taichi_three and TaichiSlAM to your dev folder and add them to PYTHONPATH

git clone https://github.com/taichi-dev/taichi_three
git clone https://github.com/xuhao1/TaichiSLAM

echo export PYTHONPATH=`pwd`/taichi_three:`pwd`/TaichiSLAM:\$PYTHONPATH >> ~/.bashrc
#Or if using zshrc
echo export PYTHONPATH=`pwd`/taichi_three:`pwd`/TaichiSLAM:\$PYTHONPATH >> ~/.zshrc

Download cow_and_lady_dataset from voxblox.

Running TaichiSLAM octomap demo

python examples/TaichiSLAM_demo.py -b ~/pathto/your/bag/cow_and_lady_dataset.bag

TSDF(Voxblox)

python examples/TaichiSLAM_demo.py -m esdf -b ~/data/voxblox/cow_and_lady_dataset.bag

Use - and = key to change accuacy. Mouse to rotate the map. -h to get more help.

usage: TaichiSLAM_demo.py [-h] [-r RESOLUTION RESOLUTION] [-m METHOD] [-c] [-t] [--rviz] [-p MAX_DISP_PARTICLES] [-b BAGPATH] [-o OCCUPY_THRES] [-s MAP_SIZE MAP_SIZE] [--blk BLK]
                          [-v VOXEL_SIZE] [-K K] [-f] [--record]

Taichi slam fast demo

optional arguments:
  -h, --help            show this help message and exit
  -r RESOLUTION RESOLUTION, --resolution RESOLUTION RESOLUTION
                        display resolution
  -m METHOD, --method METHOD
                        dense mapping method: octo/esdf
  -c, --cuda            enable cuda acceleration if applicable
  -t, --texture-enabled
                        showing the point cloud's texture
  --rviz                output to rviz
  -p MAX_DISP_PARTICLES, --max-disp-particles MAX_DISP_PARTICLES
                        max output voxels
  -b BAGPATH, --bagpath BAGPATH
                        path of bag
  -o OCCUPY_THRES, --occupy-thres OCCUPY_THRES
                        thresold for occupy
  -s MAP_SIZE MAP_SIZE, --map-size MAP_SIZE MAP_SIZE
                        size of map xy,z in meter
  --blk BLK             block size of esdf, if blk==1; then dense
  -v VOXEL_SIZE, --voxel-size VOXEL_SIZE
                        size of voxel
  -K K                  division each axis of octomap, when K>2, octomap will be K**3-map
  -f, --rendering-final
                        only rendering the final state
  --record              record to C code