SoVAR

KO | EN

gitlite — search

#python #ai #go #machine-learning #rust #deep-learning #linux #docker #nodejs #macos #awesome #kubernetes #vue #ios

(★ 22)

[ASE'24] SoVAR: Build Generalizable Scenarios from Accident Reports for Autonomous Driving Testing

SoVAR Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

This repository provides the code of SoVAR.

**We've built a full-stack app with frontend and backend – come try it out!**

- **[App Version](https://github.com/meng2180/SoVAR_app)**

### The structure of the repository

---

```
The structure of folder "src" is as follows:
src
├── file
│   ├── report                      crash reports                  
│   ├── apiKey.txt                  write your openai api-key here          
│   ├── extractionPrompt.txt        prompt with linguistic patterns            
│   ├── map.txt                     road information  
│   ├── report.txt                  input of SoVAR  
│   └── result.txt                  output of gpt        
├── solver		            solvers of car actions               
├── caculator.py                    main caculator of SoVAR  
├── generateWaypoints.py            generate waypoints          
├── gpt.py                          chat with gpt
├── search.py                       adjust points       
├── simulateReport.py               entrance to simulate report                 
├── testApollo.py                   entrance to test apollo car     
├── testRandom.py                   entrance to test apollo car with random npc
└── utils.py                        some basic methods     
```

### Requirements

---

* Python version : 3.10 or higher.

* Basic dependencies : see in requirements.txt.

* Core API  : LGSVL PythonAPI.

* Car simulator : LGSVL simulator v2021.3.

* Autopilot platform : Apollo 6.0.

* Operation system : Ubuntu 21.10.

* Hardware condition (Not mandatory) :

* GPU : Intel (R) Core (TM) i7-1070K.

* Internal memory : 32GB.
  * Display : NVIDIA GeForce RTX 4070.

### Installation

---

**1. Install LGSVL**

LGSVL simulator can be installed from https://github.com/lgsvl/simulator. The tutorial is on it as well. We are using the latest version,2021.3. LGSVL has made the difficult decision to suspend active development of SVL Simulator, as of January 1, 2022. The cloud had stopped running on June 30, 2022. Therefore, we use SORA-SVL to build our own server as a replacement.SORA-SVL can be installed from https://github.com/YuqiHuai/SORA-SVL

**2. Install Apollo6.0**

1) clone source code.

```bash
$ git clone https://github.com/ApolloAuto/apollo.git
```

2) pull docker image and enter the container.

```bash
$ sudo bash ./docker/scripts/dev_start.sh
$ sudo bash ./docker/scripts/dev_into.sh
```

3) build Apollo.

```bash
sudo ./apollo.sh build
```

4) start dreamviewer.

```bash
sudo bash scripts/bootstrap.sh
```

5) bridge Apollo with LGSVL.

```bash
bash scripts/bridge.sh
```

**3. Install SoVAR**

1) clone source code.

2) install dependencies.

```bash
pip install -r requirements.txt
```

3) import LGSVL PythonAPI.

Download python-api from https://github.com/lgsvl/PythonAPI/tree/master. After download, put the package "lgsvl" under the package "src" of SoVAR.

### Experiments

---

**E1: Reconstruct 2-NPCs car accident  from an accident report.**

* Start lgsvl simulator.
* Copy the summary part of the report in XML format to report.txt.
* Run simulateReport.py.

The two cars' behaviors are the same as the report.

**E2: test 1-EGO-and-1-NPC car accident in Apollo from an accident report.**

* Start lgsvl simulator.
* Start Apollo6.0 and connect it with lgsvl simulator.
* Copy the summary part of the report in XML format to report.txt.
* Run testApollo.py.

The NPC car's behavior is same as the report while the EGO car has the same start point and destination point as the report.

**E3: test 1-EGO-and-1-random-NPC car accident in Apollo from an accident report.**

* Start lgsvl simulator.
* Start Apollo6.0 and connect it with lgsvl simulator.
* Copy the summary part of the report in XML format to report.txt.
* Run testApollo.py.

The EGO car has the same start point and destination point as the report. But the random NPC car has a random start point and destination point.

**Q** : How to run the experiments RQ1, RQ2 and RQ3 in the paper ?

Experiments above are the usage mode of SoVAR in practical application. You can rerun RQ1, RQ2 and RQ3 in the paper by modifying a few codes based on E1, E2 and E3.

## Citation

```shell
@inproceedings{10.1145/3691620.3695037,
author = {Guo, An and Zhou, Yuan and Tian, Haoxiang and Fang, Chunrong and Sun, Yunjian and Sun, Weisong and Gao, Xinyu and Luu, Anh Tuan and Liu, Yang and Chen, Zhenyu},
title = {SoVAR: Build Generalizable Scenarios from Accident Reports for Autonomous Driving Testing},
year = {2024},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3691620.3695037},
doi = {10.1145/3691620.3695037},
booktitle = {Proceedings of the 39th IEEE/ACM International Conference on Automated Software Engineering},
pages = {268–280},
numpages = {13},
location = {Sacramento, CA, USA}
}
```

SoVAR

SoVAR Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

SoVAR

SoVAR Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README