Skip to content

official code for Customizable Embodied Multi-modal Perturbations for SLAM Robustness Benchmarking

License

Notifications You must be signed in to change notification settings

Xiaohao-Xu/SLAM-under-Perturbation

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

47 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Customizable Perturbations for RGB-D SLAM Robustness Evaluation

UMich Robotics · CMU Robotics · CMU ECE


Our Benchmarking Code is Released! See Instruction 🔥 🔥 🔥

  • Generative models, such as Sora, can simulate very COOL videos but fail to capture the physics and dynamics of our Real World.

  • We highlight the uniqueness and merits of physics-aware Noisy World simulators, and propose a customizable perturbation synthesis pipeline that can transform a Clean World to a Noisy World in a controllable manner.

Pipeline Overview

Logo

  • Noisy data synthesis pipeline for SLAM evaluation under perturbation.
  • (a) Given the customizable robot system and global trajectory, (b) the local trajectory of each sensor can be generated via the physics engine. (c) Subsequently, the trajectory perturbation composer introduces deviations to simulate locomotion perturbations (d) Following this, the render combines sensor configurations, perturbed local trajectories, and 3D scene models to generate sensor streams. (e) Finally, the sensor perturbation composer introduces corruptions to the clean sensor streams, (f) resulting in perturbed data for SLAM robustness benchmarking

Abstract

  • Robustness is a crucial factor for the successful deployment of robots in unstructured environments, particularly in the domain of Simultaneous Localization and Mapping (SLAM).
  • We propose a novel, customizable pipeline for noisy data synthesis, aimed at assessing the resilience of multi-modal SLAM models against various perturbations.
  • We introduce comprehensive perturbation taxonomy along with a perturbation composition toolbox, allowing the transformation of clean simulations into challenging noisy environments.
  • Utilizing the pipeline, we instantiate a benchmark with diverse perturbation types, to evaluate the risk tolerance of top-performing RGB-D SLAM models.
  • Our extensive analysis uncovers the model-specific susceptibilities of existing SLAM models to real-world disturbance, despite their demonstrated accuracy in standard benchmarks.

Visualizations of RGB-D SLAM Models under Perturbation

😊 Successful Cases on SplaTAM-S Model

Logo

🆘 Failure Cases on SplaTAM-S Model

Logo

😊 Successful Cases on ORB-SLAM3 Model

Logo

🆘 Failure Cases on ORB-SLAM3 Model

Logo

More Directions to Explore

  • Perturbation: Evaluate the SLAM model’s robustness under mixed perturbations and more diverse perturbation types.
  • Synthesis: How to generate more realistic perturabtions and environments for more reliable SLAM simulation?
  • SLAM Model: Extend the robustness evaluation to SLAM models with more diverse input modality types, e.g., LiDAR and sonar.
  • Robustness Enhancement: Can you propose a more robust SLAM model that can survive more severe and diverse perturbations?
  • Beyond SLAM: The evaluation can be easily extended to 3D reconstruction and other robotic navigation tasks.
  • (Please refer to the paper for more details 😄)
  • Let's embrace more robust and deployable SLAM models!!

Contact

If you have any question about this project, please feel free to contact [email protected]

Public Resources Used

We gratefully acknowledge the use of the following public resources in this work:

License

Our code is released under Apache License 2.0. see LICENSE.