Skip to content
Official repository for evaluating collision checking policies that use bayesian active learning techniques
Branch: master
Clone or download
Sanjiban Choudhury
Sanjiban Choudhury modiified README
Latest commit b6d0a2d May 24, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
collision_checking_policies cleaned history May 24, 2019
dataset_processing cleaned history May 24, 2019
dependencies cleaned history May 24, 2019
src cleaned history May 24, 2019
utils
LICENSE cleaned history May 24, 2019
README.md modiified README May 24, 2019
init_setup.m cleaned history May 24, 2019
install_dependencies.m cleaned history May 24, 2019

README.md

Learning Collision Checking Policies

Official repository for evaluating collision checking policies that use bayesian active learning techniques

Related Publications

  1. Near-Optimal Edge Evaluation in Explicit Generalized Binomial Graphs (NIPS 2017)
  2. Bayesian Active Edge Evaluation on Expensive Graphs (IJCAI 2018)

Datasets

The repository containing datasets is graph_collision_checking_dataset

Setup

  1. Clone the repository and the datasets folder
  2. cd matlab_learning_collision_checking
  3. Edit init_setup.m to add the global path to the datasets folder by editing setenv('collision_checking_dataset_folder','/path/to/data')
  4. Run install_dependencies.m
  5. Run init_setup.m

Executing the algorithms

Run the file src/benchmark_coll_check_policy.m to execute the algorithms in the paper on the datasets

Creating 2D datasets

  1. Go to dataset_processing/2D_dataset_creation/
  2. Run any one of the scripts from example_environments/ to generate a set of environments corresponding to some world distribution. You may have to create a set of empty folders for the scripts to save stuff in.
  3. Run create_graph.m. This will create a 2D RGG, start and goal and save this.
  4. Run collision_check_graph.m. This will collision check the graph on a given dataset.
You can’t perform that action at this time.