EPIBT

KO | EN

gitlite — search

#javascript #typescript #llm #rust #linux #docker #nodejs #macos #java #awesome #pytorch #security #ios #open-source

(★ 9)

No description available.

EPIBT Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Enhancing PIBT via Multi-Action Operations

The code repository of the paper "Enhancing PIBT via Multi-Action Operations"(AAAI-26).

## Tests

The tests are in the [tests](tests) folder. The tests store 5 maps: `random` (random-32-32-20), `city` (Paris-1-256), `game` (brc202d), `sortation`, `warehouse`. Each map contains the map itself `map.txt`, a list of agents `agents_*.csv`, a list of tasks with a different number of goals `tasks.csv`, and with one goal `tasks_one.csv`.

| Map | Size | \|V\| | Agents num|
|-----|------|---|-----------|
|random-32-32-20|32x32|819|100, 200, ..., 800|
|Paris-1-256|256x256|47240|1000, 2000, ..., 10000|
|brc202d|481x530|43151|500, 1000, ..., 5000|
|sortation|140x500|54320|1000, 2000, ..., 10000|
|warehouse|140x500|38586|1000, 2000, ..., 10000|

## Logs

Experiment logs are located [here](https://github.com/Straple/EPIBT/tree/metrics).

## Building

You need [CMake](https://cmake.org/) (≥v3.16), [Boost](https://www.boost.org/) (≥v1.49.0)\
The code is written in C++(20).\
To build, use [CMakeLists.txt](CMakeLists.txt)\
There you can set up the project:
- `ENABLE_ASSERT` — enabling asserts to ensure the correctness of the project. If you are sure about it, you can disable it to speed up the project by several times.
- `ENABLE_ROTATE_MODEL` — enabling rotation action model (actions: F, R, C, w), without it, the model of movement without turns is used (actions: E, S, W, N, w).
- `THREADS_NUM` — the maximum number of threads that a project can use.
- `EPIBT_DEPTH` — the depth of operations of the EPIBT-like algorithms.
- `ENABLE_EPIBT_IO` — enabling inheritance of operations in the EPIBT-like algorithms.
- `EPIBT_REVISIT` — the maximum number of visits for each agent in the EPIBT-like algorithms.

Creates the `bin/main` executable file.

## Usage

### Example
```sh
./bin/main example.config
```

### Configuration
In the argument, you pass the path to the configuration file. It has the following settings:
- `map_type` — `RANDOM`, `CITY`, `GAME`, `SORTATION`, `WAREHOUSE`.
- `planner_type` — `PIBT`, `CAUSAL_PIBT`, `PIBT_TF` (Causal PIBT+traffic flow), `EPIBT`, `EPIBT_LNS`, `PEPIBT_LNS` (parallel `EPIBT_LNS`).
- `scheduler_type` — `CONST` (а constant sequence of tasks for each agent), `GREEDY` (stores a pool of available tasks of size 1.5*number of agents, using a greedy algorithm, it gives each agent the nearest available task.).
- `graph_guidance_type` — `ENABLE` (enables Graph Guidance Optimization)
- `map_file` — the path to the file `map.txt` where the map is stored.
- `tasks_path` — the path to the tasks file.
- `agents_path` — the path to the agent file/folder. If it's going to be a folder, then it will read the `agents_*.csv` files and in this case, `THREADS_NUM_DEFAULT` threads will solve instances with different numbers of agents in parallel.
- `steps_num` — number of steps.
- `step_time` — time limit in ms for the path planner. It may exceed this value, but not much.
- `output_path` — where to record the logs of the solution. For each test instance with the test_id number, it will create a `test_id/` folder. He will write the files there.
  - `heatmap_*.csv`, where * is all, F, C, R, w (for rotation action model), otherwise all, E, S, W, N, w. Each such file contains a table where the cell indicates the number of events when such an action occurred or -1 if it is an impassable cell of the map. With this, you can see where agents often move, where wait a lot, and even visualize it.
  - `log.csv` — contains information about each step.
  - `metrics.csv` — contains total information about instance.
    |metric|value|
    |-|-|
    |task type|LMAPF-T|
    |map type|RANDOM|
    |test id|0|
    |scheduler type|CONST|
    |planner type|EPIBT(3)|
    |graph guidance type|disable|
    |agents num|100|
    |steps num|1000|
    |finished tasks|2788|
    |throughput|2.788|
    |F|71.686|
    |R|14.751|
    |C|10.386|
    |w|3.177|
    |avg step time (ms)|0.188965|
    |avg scheduler time (ms)|0.000979093|
    |avg planner time (ms)|0.158248|

### Scripts

- `src/scripts/tester.py` — a script to run bin/main on all tests and write results to the appropriate directories.
- `src/scripts/build_metrics.py` — a script for collecting information from `metrics.csv` files inside a given folder and creating a table.
- `src/scripts/build_metrics_plot.py` — a script for visualization a tables of metrics with different algorithms.
  <img width="3668" height="1506" alt="image" src="https://github.com/user-attachments/assets/3cee76a9-d24b-4583-b628-47bb4f924233" />
- `src/scripts/build_heatmap.py` — a script for visualizing a heat map of agents actions.
  <img width="2103" height="1865" alt="image" src="https://github.com/user-attachments/assets/1387fd2e-92a7-42a1-a1c9-4998d9364913" />
- `src/scripts/build_spider_plot.py` — a script for visualization a spider plot of algorithms.
  <img width="2024" height="1833" alt="image" src="https://github.com/user-attachments/assets/b7bac381-1bdf-40c9-bf27-9759a7c38eda" />

## License

This software is released under the MIT License, see [LICENSE](LICENSE).

If possible, please consider contacting the author for commercial use. This is not a restriction, I just want to know about the use as an academic researcher.

EPIBT

EPIBT Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

EPIBT

EPIBT Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README