Update README.md
Browse files
README.md
CHANGED
|
@@ -39,6 +39,14 @@ The model was trained using a hybrid approach:
|
|
| 39 |
- Target Update Frequency: Every 100 episodes
|
| 40 |
- Number of Episodes: 50
|
| 41 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
## Usage
|
| 43 |
|
| 44 |
To use this model, load the saved state dictionary and initialize the DQN with the same architecture. The model can then be used to navigate a floorplan and find the most efficient path to the target.
|
|
@@ -84,6 +92,21 @@ state = ... # Define your state here
|
|
| 84 |
with torch.no_grad():
|
| 85 |
action = model(torch.tensor(state, dtype=torch.float32).unsqueeze(0)).argmax().item()
|
| 86 |
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 87 |
## Evaluation
|
| 88 |
|
| 89 |
The model was evaluated based on:
|
|
@@ -104,6 +127,10 @@ The model was evaluated based on:
|
|
| 104 |
|
| 105 |
This project leverages the power of reinforcement learning combined with traditional pathfinding algorithms to navigate complex environments efficiently.
|
| 106 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 107 |
## Citation
|
| 108 |
|
| 109 |
If you use this model in your research, please cite it as follows:
|
|
@@ -112,7 +139,7 @@ If you use this model in your research, please cite it as follows:
|
|
| 112 |
author = {Christopher Jones},
|
| 113 |
title = {Deep Q-Network for Floorplan Navigation},
|
| 114 |
year = {2024},
|
| 115 |
-
howpublished = {\url{https://huggingface.co/cajcodes/dqn-floorplan-
|
| 116 |
note = {Accessed: YYYY-MM-DD}
|
| 117 |
}
|
| 118 |
```
|
|
|
|
| 39 |
- Target Update Frequency: Every 100 episodes
|
| 40 |
- Number of Episodes: 50
|
| 41 |
|
| 42 |
+
## Checkpoints
|
| 43 |
+
|
| 44 |
+
Checkpoints are saved during training for convenience:
|
| 45 |
+
- `checkpoint_11.pth.tar`: After 11 episodes
|
| 46 |
+
- `checkpoint_21.pth.tar`: After 21 episodes
|
| 47 |
+
- `checkpoint_31.pth.tar`: After 31 episodes
|
| 48 |
+
- `checkpoint_41.pth.tar`: After 41 episodes
|
| 49 |
+
|
| 50 |
## Usage
|
| 51 |
|
| 52 |
To use this model, load the saved state dictionary and initialize the DQN with the same architecture. The model can then be used to navigate a floorplan and find the most efficient path to the target.
|
|
|
|
| 92 |
with torch.no_grad():
|
| 93 |
action = model(torch.tensor(state, dtype=torch.float32).unsqueeze(0)).argmax().item()
|
| 94 |
```
|
| 95 |
+
## Training Script
|
| 96 |
+
|
| 97 |
+
The training script train.py is included in the repository for those who wish to reproduce the training process or continue training from a specific checkpoint.
|
| 98 |
+
|
| 99 |
+
### Training Instructions
|
| 100 |
+
- Clone the repository.
|
| 101 |
+
- Ensure you have the necessary dependencies installed.
|
| 102 |
+
- Run the training script:
|
| 103 |
+
```
|
| 104 |
+
bash
|
| 105 |
+
Copy code
|
| 106 |
+
python train.py
|
| 107 |
+
```
|
| 108 |
+
To continue training from a checkpoint, modify the script to load the checkpoint before training.
|
| 109 |
+
|
| 110 |
## Evaluation
|
| 111 |
|
| 112 |
The model was evaluated based on:
|
|
|
|
| 127 |
|
| 128 |
This project leverages the power of reinforcement learning combined with traditional pathfinding algorithms to navigate complex environments efficiently.
|
| 129 |
|
| 130 |
+
## License
|
| 131 |
+
|
| 132 |
+
This model is licensed under the Apache 2.0 License.
|
| 133 |
+
|
| 134 |
## Citation
|
| 135 |
|
| 136 |
If you use this model in your research, please cite it as follows:
|
|
|
|
| 139 |
author = {Christopher Jones},
|
| 140 |
title = {Deep Q-Network for Floorplan Navigation},
|
| 141 |
year = {2024},
|
| 142 |
+
howpublished = {\url{https://huggingface.co/cajcodes/dqn-floorplan-navigator}},
|
| 143 |
note = {Accessed: YYYY-MM-DD}
|
| 144 |
}
|
| 145 |
```
|