README.md 678 字节
Newer Older
L
LI Yunxiang 已提交
1 2 3 4 5 6 7 8
# LiftSim基线

## 简介

基于PARL库实现Deep Q-network算法,应用于[RLSchool][rlschool]库中的电梯调度模拟环境[LiftSim][liftsim]

## 依赖库

H
Hongsheng Zeng 已提交
9 10 11
+ [paddlepaddle==1.5.1](https://github.com/PaddlePaddle/Paddle)
+ [parl==1.1.2](https://github.com/PaddlePaddle/PARL)
+ [rlschool>=0.0.1](rlschool)
L
LI Yunxiang 已提交
12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

## 运行

```python
python demo.py
```

## Benchmark

<img src="rl_10.png" width="400"/>

Accumulated Reward:每3600 steps内reward的总和,可体现电梯调度在单位时间(模拟环境0.5小时)内的效率。

[rlschool]: https://github.com/PaddlePaddle/RLSchool
[liftsim]: https://github.com/PaddlePaddle/RLSchool/tree/master/rlschool/liftsim