Amigoshan / tut20_people_avoidance

tut20_people_avoidance

Dependencies

• Python package pip install msgpack-rpc-python numpy opencv-python

• PyTorch Version 0.5.0+, version 0.5.0, 1.3.0 are tested.

• Simulation environment

  Download the binary file from: [CLUSTER]/data/datasets/wenshanw/unreal/Blocks_tut20.

Test existing DQN model

• Open the simulation environment sh Blocks.sh -windowed

• Run the model: python dqn_people_avoidance.py --use-int-plotter --omni-dir --multi-frame 3 --test --load-qnet --qnet-model 4_3_dqn_2000000.pkl

Train a new model

• Open the simulation environment

• Run the command: python dqn_people_avoidance.py --exp-prefix "1_7_" --use-int-plotter --train-step 2000000 --train-interval 100 --batch-size 1000 --lr 0.001 --omni-dir --multi-frame 3