Implementation of Model Predictive Motion Planning Network
cd into deps/ and build cpp modules based on sparse-rrt python binding.
[Important]Please make sure all sys.path inside scripts you need have been updated to your own version. Such as benchmark.py and scripts inside expriments
The data_gen folder contains code and scripts for four envrionments: acrobot, cart-pole, car and quadrotor.
cd data_gen
# For parallelization, run the script in i different process:
bash datagen_${system}_batch${i}.shPreprocess obstacles and state-goal paris with process_data.py, process_obs.py in mpnet/sst_envs Train the network with mpnet/train_mpnet.py.
For costnet in MPC-MPNet-Path, use mpnet/train_costs.py
Editing parmas in params folder and run with script
# example: bash scripts/acrobot_obs/mp_tree.sh
bash scripts/${system}/${method}.shIf you find this open source release useful, please reference in your paper:
@misc{2101.06798,
Author = {Linjun Li and Yinglong Miao and Ahmed H. Qureshi and Michael C. Yip},
Title = {MPC-MPNet: Model-Predictive Motion Planning Networks for Fast, Near-Optimal Planning under Kinodynamic Constraints},
Year = {2021},
Eprint = {arXiv:2101.06798},
}