IR-MCL: Implicit Representation-Based Online Global Localization

<a href="https://www.ipb.uni-bonn.de/people/haofei-kuang/"><strong>Haofei Kuang</strong></a>
·
<a href="https://www.ipb.uni-bonn.de/people/xieyuanli-chen/"><strong>Xieyuanli Chen</strong></a>
·
<a href="https://phd.uniroma1.it/web/TIZIANO-GUADAGNINO_nP1536210_IT.aspx"><strong>Tiziano Guadagnino</strong></a>
·
<a href="https://www.ipb.uni-bonn.de/people/nicky-zimmerman/"><strong>Nicky Zimmerman</strong></a>
·
<a href="https://www.ipb.uni-bonn.de/people/jens-behley/"><strong>Jens Behley</strong></a>
·
<a href="https://www.ipb.uni-bonn.de/people/cyrill-stachniss/"><strong>Cyrill Stachniss</strong></a>

University of Bonn

Online localization demo

Abstract

Determining the state of a mobile robot is an essential building block of robot navigation systems. In this paper, we address the problem of estimating the robot’s pose in an indoor environment using 2D LiDAR data and investigate how modern environment models can improve gold standard Monte-Carlo localization (MCL) systems. We propose a neural occupancy field (NOF) to implicitly represent the scene using a neural network. With the pretrained network, we can synthesize 2D LiDAR scans for an arbitrary robot pose through volume rendering. Based on the implicit representation, we can obtain the similarity between a synthesized and actual scan as an observation model and integrate it into an MCL system to perform accurate localization. We evaluate our approach on five sequences of a self-recorded dataset and three publicly available datasets. We show that we can accurately and efficiently localize a robot using our approach surpassing the localization performance of state-of-the-art methods. The experiments suggest that the presented implicit representation is able to predict more accurate 2D LiDAR scans leading to an improved observation model for our particle filter-based localization.

Dependencies

The code was tested with Ubuntu 20.04 with: - python version 3.9. - pytorch version 1.13.1 with CUDA 11.6 - pytorch-lighting with 1.9.0

Installation

• Clone the repo: shell git clone https://github.com/PRBonn/ir-mcl.git cd ir-mcl

• Prepare the python environment (Anaconda is recommended here): shell conda env create -f environment.yml or ```shell conda create --name irmcl python=3.9.13 conda activate irmcl

conda install -c conda-forge pybind11 pip install torch torchvision --extra-index-url https://download.pytorch.org/whl/cu116 pip install pytorch-lightning tensorboardX pip install matplotlib scipy open3d pip install evo --upgrade --no-binary evo ```

• Compile the motion model and resampling module shell cd ir-mcl/mcl & conda activate ir-mcl make -j4

Preparation

Datasets

Please refer to PREPARE_DATA to prepare the datasets

Pre-trained Weights

The pre-trained weights are stored at config folder, includes: - IPBLab dataset: config/ipblab_nof_weights.ckpt - Freiburg Building 079 dataset: config/fr079_nof_weights.ckpt - Intel Lab dataset: config/intel_nof_weights.ckpt - MIT CSAIL dataset: config/mit_nof_weights.ckpt

Run Experiments

Global Localization Experiments on IPBLab dataset

• Pre-training NOF on IPBLab dataset (The train/eval/test set of IPBLab dataset are not available now, they will be released after our dataset paper is published!) shell cd ~/ir-mcl bash ./shells/pretraining/ipblab.sh

• Global localization experiments shell cd ir-mcl python main.py --config_file ./config/global_localization/loc_config_{sequence_id}.yml # for example: python main.py --config_file ./config/global_localization/loc_config_test1.yml

• Pose-tracking experiments shell cd ir-mcl python main.py --config_file ./config/pose_tracking/loc_config_{sequence_id}.yml # for example: python main.py --config_file ./config/pose_tracking/loc_config_test1.yml

Observation Model Experiments

• Train/Test (replace "dataset" in "fr079", "intel", or "mit") shell cd ir-mcl bash ./shells/pretraining/{dataset}.sh # for example: bash ./shells/pretraining/intel.sh

Supplements for the Experimental Results

Due to the space limitation of the paper, we provide some experimental results as supplements here.

Memery cost

We provide an ablation study on the memory cost between the occupancy grid map (OGM), Hilbert map, and our neural occupancy field (NOF).

Maps type	Approximate memory	Loc. method	RMSE: location (cm) / yaw (degree)
OGM (5cm grid size)	4.00MB	AMCL

NMCL

SRRG-Loc | 11.11 / 4.15

19.57 / 3.62

8.74 / 1.68 | | OGM (10cm grid size) | 2.00MB | AMCL

NMCL

SRRG-Loc | 15.01 / 4.18

36.27 / 4.04

12.15 / 1.53 | | Hilbert Map | 0.01MB | HMCL | 20.04 / 4.50 | | NOF | 1.96NB | IR-MCL | 6.62 / 1.11 |

Ablation study on fixed particle numbers

We also provide the experiment to study the performance of global localization under the same particle numbers for all methods. We fixed the number of particles to 100,000. In the below table, all baselines and IR-MCL^∗ always use 100,000 particles. IR-MCL is shown for reference.

Method	RMSE: location (cm) / yaw (degree)
AMCL

NMCL

HMCL

SRRG-Loc

IR-MCL^∗ | 11.56 / 4.12

19.57 / 3.62

20.54 / 4.70

8.74 / 1.68

6.71 / 1.11 | | IR-MCL | 6.62 / 1.11 |

Citation

If you use this library for any academic work, please cite our original paper.

@article{kuang2023ral,
  author    = {Kuang, Haofei and Chen, Xieyuanli and Guadagnino, Tiziano and Zimmerman, Nicky and Behley, Jens and Stachniss, Cyrill},
  title     = {{IR-MCL: Implicit Representation-Based Online Global Localization}},
  journal   = {IEEE Robotics and Automation Letters (RA-L)},
  doi       = {10.1109/LRA.2023.3239318},
  year      = {2023},
  codeurl   = {https://github.com/PRBonn/ir-mcl},
}

Acknowledgment

This work has partially been funded by the European Union’s Horizon 2020 research and innovation programme under grant agreement No 101017008 (Harmony).