> For the complete documentation index, see [llms.txt](https://2025.istvs.org/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://2025.istvs.org/submissions/papers/2824.md).

# 2824 - LOCAL PATH PLANNING ON ROUGH TERRAIN FOR UNMANNED GROUND VEHICLE BY REINFORCEMENT LEARNING CONSIDERING SEARCH REWARD

{% hint style="info" %}
Paper presented at ISTVS 2025 | 55th Conference of the International Society for Terrain-Vehicle Systems <https://doi.org/10.56884/HVHA4780>
{% endhint %}

**Authors:** *Ryosuke Eto, Fushiki Taguchi, Junya Yamakawa*

**Keywords:** Local path planning; autonomous unmanned ground vehicle; reinforcement learning; rough terrain

**Abstract:**

Autonomous unmanned ground vehicles are expected to be utilized at disaster sites. The previously obtained map information for the site is often unreliable because disaster areas can change rapidly and unpredictably. Therefore, the vehicles must be able to recognize the surrounding rough terrain environment in real time and plan optimal routes to approach the destination safely. However, local path planning may fall into the local minimum solution and get stuck. Therefore, in this study, a method of avoiding a standstill by utilizing information entropy and route history to encourage exploratory behavior was investigated. In the proposed method, routes are selected by determining the vehicle behavior using reinforcement learning. For observation information, a Digital Elevation Model, an information entropy map of elevation, and a map that records the number of times the vehicle has traveled on the grid, in addition to the vehicle's state were used. The aim is to acquire the behavior of going to the target with selecting a safe route to avoid getting stuck by rewarding the vehicle for going to the target, avoiding uneven terrain, reducing information entropy, and not repeatedly driving in the same place. In this study, we tested the proposed method on each terrain with a single large obstacle and a dead-end by simulations. The simulation results showed that the vehicle can reach the target efficiently by introducing rewards for exploratory behavior.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://2025.istvs.org/submissions/papers/2824.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
