MarsSimNav: Autonomous Terrain-Aware Rover Path Planning Overview

MarsSimNav is a deep learning and path planning system built to simulate autonomous Mars rover navigation using real NASA imagery. It leverages a DeepLabV3+ segmentation model trained on the AI4Mars dataset to classify Martian terrain types, followed by A* search for optimal path planning over hazardous terrain. Features

Semantic Segmentation with DeepLabV3+

Custom PyTorch Dataset Pipeline for AI4Mars

Terrain-Aware Path Planning using A* algorithm

Terrain Cost Mapping: soil, bedrock, sand, and big rocks

Visual Overlay of predicted rover path on Mars images

Efficient data handling for large datasets (supports subsetting)

Inference-ready model with reproducible outputs

Dataset

Source: AI4Mars Dataset

Images: Over 18,000 images from the Curiosity rover (MSL)

Labels: Pixel-level annotations for soil, bedrock, sand, and big rocks

Preprocessing: Optionally subset to 5,000 images for lightweight experimentation

Technologies Used

Python 3.11

PyTorch and Torchvision

DeepLabV3+ with ResNet50 backbone

NetworkX for path planning (A* algorithm)

Matplotlib and PIL for visualization

Google Colab for development and testing

Getting Started

Clone the repository:

git clone https://github.com/pushkar-hue/marssimnav.git cd marssimnav

Install dependencies:

pip install -r requirements.txt

To run in Google Colab:

Upload your trained model as deeplabv3_mars.pth

Place sample Mars images in ai4mars-subset/images/

Call plan_rover_path(image_path, model) to visualize paths

Path Planning Logic

Predict segmentation mask for a Mars image using DeepLabV3+

Convert mask into a terrain cost map

Use A* search to compute the lowest-cost safe path

Visualize the path overlay on both the mask and the original image

Terrain Classes and Costs Terrain Class ID Cost Soil 0 1 Bedrock 1 4 Sand 2 6 Big Rock 3 10

Pixels with unknown labels (e.g. masked/NULL) are assigned high cost or skipped. Sample Results Input Image Segmentation Mask Planned Path Overlay mars_image.jpg predicted_mask.png planned_path_overlay.png

Future Work

Add support for D* Lite and real-time replanning

Interactive terrain selection and goal input via UI

Animated simulation of rover motion along the path

Deployable web dashboard using Streamlit or Gradio

Integration with planetary data system (PDS) metadata

Authors

Pushkar Sharma (@pushkar-hue)
Ved Thorat (@i3hz)

AI4Mars Dataset by NASA JPL and collaborators