README.md

---

language:
- en
license: mit
task_categories:
- robotics
tags:
- agent
- robotics
- benchmark
- environment
- underwater
- multi-modal
- mllm
- large-language-models
size_categories:
- n<1K
---


<h1 align="center"> 🌊 OceanGym 🦾 </h1>
<h3 align="center"> A Benchmark Environment for Underwater Embodied Agents </h3>

<p align="center">
  🌐 <a href="https://oceangpt.github.io/OceanGym" target="_blank">Home Page</a>
  📄 <a href="https://arxiv.org/abs/2509.26536" target="_blank">ArXiv Paper</a>
  🤗 <a href="https://huggingface.co/datasets/zjunlp/OceanGym" target="_blank">Hugging Face</a>
  ☁️ <a href="https://drive.google.com/file/d/1EfKHeiyQD5eoJ6-EsiJHuIdBRM5Ope5A/view?usp=drive_link" target="_blank">Google Drive</a>
  ☁️ <a href="https://pan.baidu.com/s/16h86huHLeFGAKatRWvLrFQ?pwd=wput" target="_blank">Baidu Drive</a>
</p>

  <img src="asset/img/o1.png" align=center>

**OceanGym** is a high-fidelity embodied underwater environment that simulates a realistic ocean setting with diverse scenes. As illustrated in figure, OceanGym establishes a robust benchmark for evaluating autonomous agents through a series of challenging tasks, encompassing various perception analyses and decision-making navigation. The platform facilitates these evaluations by supporting multi-modal perception and providing action spaces for continuous control.

# 💐 Acknowledgement

OceanGym environment is built upon Unreal Engine (UE) 5.3, with certain components developed by drawing inspiration from and partially based on [HoloOcean](https://github.com/byu-holoocean). We sincerely acknowledge their valuable contribution.


# 🔔 News

- 10-2025, we released the initial version of OceanGym along with the accompanying [paper](https://arxiv.org/abs/2509.26536).
- 04-2025, we launched the OceanGym project.

---

**Contents:**
- [💐 Acknowledgement](#-acknowledgement)
- [🔔 News](#-news)
- [📺 Quick Start](#-quick-start)
  - [Decision Task](#decision-task)
  - [Perception Task](#perception-task)
- [⚙️ Set up Environment](#️-set-up-environment)
  - [Clone HoloOcean](#clone-holoocean)
  - [Packaged Installation](#packaged-installation)
  - [Add World Files](#add-world-files)
  - [Open the World](#open-the-world)
- [🧠 Decision Task](#-decision-task)
  - [Target Object Locations](#target-object-locations)
  - [Evaluation Criteria](#evaluation-criteria)
- [👀 Perception Task](#-perception-task)
  - [Using the Bench to Eval](#using-the-bench-to-eval)
    - [Import Data](#import-data)
    - [Set your Model Parameters](#set-your-model-parameters)
    - [Simple Multi-views](#simple-multi-views)
    - [Multi-views with Sonar](#multi-views-with-sonar)
    - [Multi-views add Sonar Examples](#multi-views-add-sonar-examples)
  - [Collecting Image Data](#collecting-image-data)
    - [Modify Configuration File](#modify-configuration-file)
    - [Collect Camera Images Only](#collect-camera-images-only)
    - [Collect Camera and Sonar Images](#collect-camera-and-sonar-images)
- [⏱️ Results](#️-results)
  - [Decision Task](#decision-task-1)
  - [Perception Task](#perception-task-1)
- [📚 Datasets](#-datasets)
- [🚩 Citation](#-citation)

# 📺 Quick Start

Install the experimental code environment using pip:

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

## Decision Task

> Only the environment is ready! Build the environment based on [here](#️-set-up-environment).

**Step 1: Run a Task Script**

   For example, to run task 4:

   ```bash
   python decision\tasks\task4.py
   ```

   Follow the keyboard instructions or switch to LLM mode for automatic decision-making.


**Step 2: Keyboard Control Guide**

   | Key         | Action                        |
   |-------------|------------------------------|
   | W           | Move Forward                 |
   | S           | Move Backward                |
   | A           | Move Left                    |
   | D           | Move Right                   |
   | J           | Turn Left                    |
   | L           | Turn Right                   |
   | I           | Move Up                      |
   | K           | Move Down                    |
   | M           | Switch to LLM Mode           |
   | Q           | Exit                         |

   > You can use WASD for movement, J/L for turning, I/K for up/down.
   > Press `M` to switch to large language model mode (may cause temporary lag).
   > Press `Q` to exit.

**Step 3: View Results**

   Logs and memory files are automatically saved in the `log/` and `memory/` directories.

**Step 4: Evaluate the results**

   Place the generated `memory` and `important_memory` files into the corresponding `point` folders.
   Then, set the evaluation paths in the `evaluate.py` file.

   We provide 6 experimental evaluation paths. In `evaluate.py`, you can configure them as follows:

   ```python
   eval_roots = [
       os.path.join(eval_root, "main", "gpt4omini"),
       os.path.join(eval_root, "main", "gemini"),
       os.path.join(eval_root, "main", "qwen"),
       os.path.join(eval_root, "migration", "gpt4o"),
       os.path.join(eval_root, "migration", "qwen"),
       os.path.join(eval_root, "scale", "qwen"),
   ]
   ```

   To run the evaluation:

   ```bash
   python decision\utils\evaluate.py
   ```

   The generated results will be saved under the `\eval\decision` folder.

## Perception Task

> All commands are applicable to **Linux**, so if you using **Windows**, you need to change the corresponding path representation (especially the slash).

**Step 1: Prepare the dataset**

After downloading from [Hugging Face](https://huggingface.co/datasets/zjunlp/OceanGym/tree/main/data/perception) or [Google Drive](https://drive.google.com/drive/folders/1H7FTbtOCKTIEGp3R5RNsWvmxZ1oZxQih), put it into the `data/perception` folder.

**Step 2: Select model parameters**

| parameter | function |
| ---| --- |
| model_template | The large language model message queue template you selected. |
| model_name_or_path | If it is an API model, it is the model name; if it is a local model, it is the path. |
| api_key | If it is an API model, enter your key. |
| base_url | If it is an API model, enter its baseful URL. |

Now we only support OpenAI, Google Gemma, Qwen and OpenBMB.

```bash
MODELS_TEMPLATE="Yours"
MODEL_NAME_OR_PATH="Yours"
API_KEY="Yours"
BASE_URL="Yours"
```

**Step 3: Run the experiments**

| parameter | function |
| ---| --- |
| exp_name | Customize the name of the experiment to save the results. |
| exp_idx | Select the experiment number, or enter "all" to select all. |
| exp_json | JSON file containing the experiment label data. |
| images_dir | The folder where the experimental image data is stored. |

For the experimental types, We designed (1) multi-view perception task and (2) context-based perception task.

For the lighting conditions, We designed (1) high illumination and (2) low illumination.

For the auxiliary sonar, We designed (1) without sonar image (2) zero-shot sonar image and (3) sonar image with few sonar example.

Such as this command is used to evaluate the **multi-view** perception task under **high** illumination:


```bash
python perception/eval/mv.py \
    --exp_name Result_MV_highLight_00 \
    --exp_idx "all" \
    --exp_json "/data/perception/highLight.json" \
    --images_dir "/data/perception/highLight" \
    --model_template $MODELS_TEMPLATE \
    --model_name_or_path $MODEL_NAME_OR_PATH \
    --api_key $API_KEY \
    --base_url $BASE_URL
```

For more patterns about perception tasks, please read [this](#-perception-task) part carefully.

# ⚙️ Set up Environment

This project is based on the HoloOcean environment. 💐

> We have placed a simplified version here. If you encounter any detailed issues, please refer to the [original installation document](https://byu-holoocean.github.io/holoocean-docs/v2.1.0/usage/installation.html).


## Install the OceanGym_large.zip 

From 
  ☁️ <a href="https://drive.google.com/file/d/1EfKHeiyQD5eoJ6-EsiJHuIdBRM5Ope5A/view?usp=drive_link" target="_blank">Google Drive</a>
  ☁️ <a href="https://pan.baidu.com/s/16h86huHLeFGAKatRWvLrFQ?pwd=wput" target="_blank">Baidu Drive</a>
  download the **OceanGym_large.zip** And extract it to the folder you want
  
## Packaged Installation

1. Python Library

From the cloned repository, install the Python package by doing the following:

```bash
cd OceanGym_large/client
pip install .
```

2. Worlds Packages

Install the package by running the following Python commands:

```python
import holoocean
holoocean.install("Ocean")
```

To do these steps in a single console command, use:

```bash
python -c "import holoocean; holoocean.install('Ocean')"
```

## Add World Files

Place the JSON config file from `asset/decision/map_config` or `asset\perception\map_config` into some place like:

(Windows)

```
C:\Users\Windows\AppData\Local\holoocean\2.0.0\worlds\Ocean
```

## Open the World

**1. If you're use it in first time, you have to compile it**

  1-1. find the Holodeck.uproject in **engine** folder
  
  <img src="asset/img/pic1.png" style="width: 60%; height: auto;" align="center">

  1-2. Right-click and select:Generate Visual Studio project files
  
  <img src="asset/img/pic2.png" style="width: 60%; height: auto;" align="center">

  1-3. If the version is not 5.3.2,please choose the Switch Unreal Engine Version
  
  <img src="asset/img/pic3.png" style="width: 60%; height: auto;" align="center">

  1-4. Then open the project
  
  <img src="asset/img/pic4.png" style="width: 60%; height: auto;" align="center">

**2. Then find the `HAIDI` map in `demo` directory**

  <img src="asset/img/pic5.png" style="width: 60%; height: auto;" align="center">

**3. Run the project**

  <img src="asset/img/pic6.png" style="width: 60%; height: auto;" align="center">

**4. Run the code**

 When the ue editor shows as follows, namely: **"LogD3D12RHI: Cannot end block when stack is empty"** , it indicates that the scene has been loaded. 
 
  <img src="asset/img/pic7.png" style="width: 60%; height: auto;" align="center">
  
Then you can start the code, either directly using vscode 

 <img src="asset/img/pic8.png" style="width: 60%; height: auto;" align="center">
or by entering the following command in the command line

```bash
python decision\tasks\task4.py
```

# 🧠 Decision Task

> All commands are applicable to **Windows** only, because it requires full support from the `UE5 Engine`.

The decision experiment can be run with reference to the [Quick Start](#-quick-start).

## Target Object Locations

We have provided eight tasks. For specific task descriptions, please refer to the [paper](https://arxiv.org/abs/2509.26536).

The following are the coordinates for each target object in the environment (in meters):

- **MINING ROBOT**:
  (-71, 149, -61), (325, -47, -83)
- **OIL PIPELINE**:
  (345, -165, -32), (539, -233, -42), (207, -30, -66)
- **OIL DRUM**:
  (447, -203, -98)
- **SUNKEN SHIP**:
  (429, -151, -69), (78, -11, -47)
- **ELECTRICAL BOX**:
  (168, 168, -65)
- **WIND POWER STATION**:
  (207, -30, -66)
- **AIRCRAFT WRECKAGE**:
  (40, -9, -54), (296, 78, -70), (292, -186, -67)
- **H-MARKED LANDING PLATFORM**:
  (267, 33, -80)

---

## Evaluation Criteria

1. If the target is not found, use the final stopping position for evaluation.
2. If the target is found, use the closest distance to any target point.
3. For found targets:
   - Minimum distance ≤ 30: full score
   - 30 < distance < 100: score decreases proportionally
   - Distance ≥ 100: score is 0
4. Score composition:
   - One point: 100
   - Two points: 60 / 40
   - Three points: 60 / 20 / 20

# 👀 Perception Task

## Using the Bench to Eval

> All commands are applicable to **Linux**, so if you using **Windows**, you need to change the corresponding path representation (especially the slash).
>
> Now we only support OpenAI, Google Gemma, Qwen and OpenBMB. If you need to customize the model, please contact the author.

### Import Data

First, you need download our data from [Hugging Face](https://huggingface.co/datasets/zjunlp/OceanGym) or [Google Drive](https://drive.google.com/drive/folders/1H7FTbtOCKTIEGp3R5RNsWvmxZ1oZxQih).

And then create a new `data` folder in the project root directory:

```bash
mkdir -p data/perception
```

Finally, put the downloaded data into the corresponding folder.

### Set your Model Parameters

Just open a terminal in the root directory and set it directly.

| parameter | function |
| ---| --- |
| model_template | The large language model message queue template you selected. |
| model_name_or_path | If it is an API model, it is the model name; if it is a local model, it is the path. |
| api_key | If it is an API model, enter your key. |
| base_url | If it is an API model, enter its baseful URL. |

```bash
MODELS_TEMPLATE="Yours"
MODEL_NAME_OR_PATH="Yours"
API_KEY="Yours"
BASE_URL="Yours"
```

### Simple Multi-views

All of these scripts evaluate the perception task, and the parameters are as follows:

| parameter | function |
| ---| --- |
| exp_name | Customize the name of the experiment to save the results. |
| exp_idx | Select the experiment number, or enter "all" to select all. |
| exp_json | JSON file containing the experiment label data. |
| images_dir | The folder where the experimental image data is stored. |

This command is used to evaluate the **multi-view** perception task under **high** illumination:

```bash
python perception/eval/mv.py \
    --exp_name Result_MV_highLight_00 \
    --exp_idx "all" \
    --exp_json "/data/perception/highLight.json" \
    --images_dir "/data/perception/highLight" \
    --model_template $MODELS_TEMPLATE \
    --model_name_or_path $MODEL_NAME_OR_PATH \
    --api_key $API_KEY \
    --base_url $BASE_URL
```

This command is used to evaluate the **context-based** perception task under **high** illumination:

```bash
python perception/eval/mv.py \
    --exp_name Result_MV_highLightContext_00 \
    --exp_idx "all" \
    --exp_json "/data/perception/highLightContext.json" \
    --images_dir "/data/perception/highLightContext" \
    --model_template $MODELS_TEMPLATE \
    --model_name_or_path $MODEL_NAME_OR_PATH \
    --api_key $API_KEY \
    --base_url $BASE_URL
```

This command is used to evaluate the **multi-view** perception task under **low** illumination:

```bash
python perception/eval/mv.py \
    --exp_name Result_MV_lowLight_00 \
    --exp_idx "all" \
    --exp_json "/data/perception/lowLight.json" \
    --images_dir "/data/perception/lowLight" \
    --model_template $MODELS_TEMPLATE \
    --model_name_or_path $MODEL_NAME_OR_PATH \
    --api_key $API_KEY \
    --base_url $BASE_URL
```

This command is used to evaluate the **context-based** perception task under **low** illumination:

```bash
python perception/eval/mv.py \
    --exp_name Result_MV_lowLightContext_00 \
    --exp_idx "all" \
    --exp_json "/data/perception/lowLightContext.json" \
    --images_dir "/data/perception/lowLightContext" \
    --model_template $MODELS_TEMPLATE \
    --model_name_or_path $MODEL_NAME_OR_PATH \
    --api_key $API_KEY \
    --base_url $BASE_URL
```

### Multi-views with Sonar

This command is used to evaluate the **multi-view** perception task under **high** illumination with **sonar** image:

```bash
python perception/eval/mvs.py \
    --exp_name Result_MVwS_highLight_00 \
    --exp_idx "all" \
    --exp_json "/data/perception/highLight.json" \
    --images_dir "/data/perception/highLight" \
    --model_template $MODELS_TEMPLATE \
    --model_name_or_path $MODEL_NAME_OR_PATH \
    --api_key $API_KEY \
    --base_url $BASE_URL
```

This command is used to evaluate the **context-based** perception task under **high** illumination with **sonar** image:

```bash
python perception/eval/mvs.py \
    --exp_name Result_MVwS_highLightContext_00 \
    --exp_idx "all" \
    --exp_json "/data/perception/highLightContext.json" \
    --images_dir "/data/perception/highLightContext" \
    --model_template $MODELS_TEMPLATE \
    --model_name_or_path $MODEL_NAME_OR_PATH \
    --api_key $API_KEY \
    --base_url $BASE_URL
```

This command is used to evaluate the **multi-view** perception task under **low** illumination with **sonar** image:

```bash
python perception/eval/mvs.py \
    --exp_name Result_MVwS_lowLight_00 \
    --exp_idx "all" \
    --exp_json "/data/perception/lowLight.json" \
    --images_dir "/data/perception/lowLight" \
    --model_template $MODELS_TEMPLATE \
    --model_name_or_path $MODEL_NAME_OR_PATH \
    --api_key $API_KEY \
    --base_url $BASE_URL
```

This command is used to evaluate the **context-based** perception task under **low** illumination with **sonar** image:

```bash
python perception/eval/mvs.py \
    --exp_name Result_MVwS_lowLightContext_00 \
    --exp_idx "all" \
    --exp_json "/data/perception/lowLightContext.json" \
    --images_dir "/data/perception/lowLightContext" \
    --model_template $MODELS_TEMPLATE \
    --model_name_or_path $MODEL_NAME_OR_PATH \
    --api_key $API_KEY \
    --base_url $BASE_URL
```

### Multi-views add Sonar Examples

This command is used to evaluate the **multi-view** perception task under **high** illumination with **sona** image **examples**:

```bash
python perception/eval/mvsex.py \
    --exp_name Result_MVwSss_highLight_00 \
    --exp_idx "all" \
    --exp_json "/data/perception/highLight.json" \
    --images_dir "/data/perception/highLight" \
    --model_template $MODELS_TEMPLATE \
    --model_name_or_path $MODEL_NAME_OR_PATH \
    --api_key $API_KEY \
    --base_url $BASE_URL
```

This command is used to evaluate the **context-based** perception task under **high** illumination with **sona** image **examples**:

```bash
python perception/eval/mvsex.py \
    --exp_name Result_MVwSss_highLightContext_00 \
    --exp_idx "all" \
    --exp_json "/data/perception/highLightContext.json" \
    --images_dir "/data/perception/highLightContext" \
    --model_template $MODELS_TEMPLATE \
    --model_name_or_path $MODEL_NAME_OR_PATH \
    --api_key $API_KEY \
    --base_url $BASE_URL
```

This command is used to evaluate the **multi-view** perception task under **low** illumination with **sona** image **examples**:

```bash
python perception/eval/mvsex.py \
    --exp_name Result_MVwSss_lowLight_00 \
    --exp_idx "all" \
    --exp_json "/data/perception/lowLight.json" \
    --images_dir "/data/perception/lowLight" \
    --model_template $MODELS_TEMPLATE \
    --model_name_or_path $MODEL_NAME_OR_PATH \
    --api_key $API_KEY \
    --base_url $BASE_URL
```

This command is used to evaluate the **context-based** perception task under **low** illumination with **sona** image **examples**:

```bash
python perception/eval/mvsex.py \
    --exp_name Result_MVwSss_lowLightContext_00 \
    --exp_idx "all" \
    --exp_json "/data/perception/lowLightContext.json" \
    --images_dir "/data/perception/lowLightContext" \
    --model_template $MODELS_TEMPLATE \
    --model_name_or_path $MODEL_NAME_OR_PATH \
    --api_key $API_KEY \
    --base_url $BASE_URL
```

## Collecting Image Data

> This part is optional. Only use when you need to collect pictures by yourself.

### Modify Configuration File

The sample configuration files can be found in `asset/perception/map_config`. You need to copy this and paste it into your HoloOcean project's configuration.

### Collect Camera Images Only

This command is used to collect **camera** images only, and the parameters are as follows:

| parameter | function |
| ---| --- |
| scenario | The name of the json configuration file you want to replace. |
| task_name | Customize the name of the experiment to save the results. |
| rgbcamera | The camera directions you can choose. If select all, enter "all". |

```bash
python perception/task/init_map.py \
    --scenario without_sonar \
    --task_name "Exp_Camera_Only" \
    --rgbcamera "all"
```

### Collect Camera and Sonar Images

This command is used to collect both **camera** images and **sonar** images at same time:

```bash
python perception/task/init_map_with_sonar.py \
    --scenario with_sonar \
    --task_name "Exp_Add_Sonar" \
    --rgbcamera "FrontCamera"
```

# ⏱️ Results

**We provide the trajectory data of OceanGym’s various task evaluations at the [next section](#-datasets), enabling readers to analyze and reproduce the results.**

## Decision Task

  <img src="asset/img/t1.png" align=center>

- This table is the performance in decision tasks requiring autonomous completion by MLLM-driven agents.

## Perception Task

  <img src="asset/img/t2.png" align=center>

- This table is the performance of perception tasks across different models and conditions.
- Values represent accuracy percentages.
- Adding sonar means using both RGB and sonar images.

# 📚 DataSets
**The link to the dataset is as follows**\
 ☁️ <a href="https://drive.google.com/drive/folders/1VhrvhvbWvnaS4EyeyaV1fmTQ6gPo8GCN?usp=drive_link" target="_blank">Google Drive</a>
- Decision Task

```
decision_dataset
├── main
│ ├── gpt4omini
│ │ ├── task1
│ │ │ ├── point1
│ │ │ │ ├── llm_output_...log
│ │ │ │ ├── memory_...json
│ │ │ │ └── important_memory_...json
│ │ │ └── ... (other data points like point2, point3...)
│ │ └── ... (other tasks like task2, task3...)
│ ├── gemini
│ │ └── ... (structure is the same as gpt4omini)
│ └── qwen
│ └── ... (structure is the same as gpt4omini)
│
├── migration
│ ├── gpt4o
│ │ └── ... (structure is the same as above)
│ └── qwen
│ └── ... (structure is the same as above)
│
└── scale
  ├── qwen
  └── gpt4omini
```
### **How to use this dataset**

In the main folder, you can see the data generated by the three models corresponding to the three folders. Within each model folder, there are task1-12 task folders, and within the task folders, there are point1-3 folders, representing the results generated from different starting points. Among them, point1 and point2 are **fixed starting points**, which are respectively [144 ,-114,-63] and [350 ,-118 -7] and point3 is a **random point**\
In the scale experiment, Point1-4 represent different task durations, with point1 being **1 hour**, point2 **1.5 hours**, point3 **2 hours**, and point4 **3 hours**. Note that the actual duration may vary to some extent due to the influence of large model calls, network fluctuations, and other factors\
If you want to evaluate the files generated by yourself, please place the corresponding **memory_{time_stamp}.json** and **important_memory_{time_stamp}.json** files in the corresponding folders

- Perception Task

```
perception_dataset
├── data
│ ├── highLight
│ ├── highLightContext
│ ├── lowLight
│ ├── lowLightContext
│ └── ... (label files)
│
└── result
└── ... (detail result fils)
```

### **How to use this dataset**

In the main folder, `data` is the test data of perception task, `result` is the detail results of this [table](#perception-task-1).

Below the folder `data`, there are 4 folders and 4 JSON files. Each folder contains test data for each perception task, and each JSON file is the label of its corresponding folder.

# 🔧 Develop OceanGym
OceanGym supports custom scenarios. You can freely exert yourself in the scenarios we provide!\
You can find the assets you need in the **ue5 fab Mall** and add them to OceanGym to test the exploration ability of the robot!\
Or modify parameters such as **terrain and lighting** to simulate the weather in different scenarios!

### Modify lighting

Step 1:
Find the **DirectionalLight** in outliner

Step 2:
Choose the details of **DirectionalLight**

Step 3:
Modify the data of **light** as per your requirements

 <img src="asset/img/pic9.png" style="width: 60%; height: auto;" align="center">

**Notice**\
In our paper, we simulate low-light and high-light environments, where the Intensity of light is **10.0lux** in the **high-light** environment
Intensity of light is **1.5lux** in a **low-light** environment

### Modify start position
Step 1:
Find the initial config file **OceanGym.json** in
```
C:\Users\Windows\AppData\Local\holoocean\2.0.0\worlds\Ocean
```
Step 2:
Modify the data of **location** as per your requirements

 <img src="asset/img/pic10.png" style="width: 60%; height: auto;" align="center">

 If you want to develop more functions, you can visit [the official website of holoocean](https://byu-holoocean.github.io/holoocean-docs/v2.0.1/develop/develop.html)

# 🚩 Citation

If this OceanGym paper or benchmark is helpful, please kindly cite as this:

```bibtex
@misc{xue2025oceangymbenchmarkenvironmentunderwater,
      title={OceanGym: A Benchmark Environment for Underwater Embodied Agents}, 
      author={Yida Xue and Mingjun Mao and Xiangyuan Ru and Yuqi Zhu and Baochang Ren and Shuofei Qiao and Mengru Wang and Shumin Deng and Xinyu An and Ningyu Zhang and Ying Chen and Huajun Chen},
      year={2025},
      eprint={2509.26536},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2509.26536}, 
}
```

💐 Thanks again!