Design and Development of Swarm Robots for Collaborative Planetary Exploration

Planetary exploration missions demand adaptable, resilient robotic systems due to their inherently unpredictable terrains and limited Earth communication. Traditional singlerover systems are limited in scalability, redundancy, and real-time decision-making. This work presents the design and partial realization of a decentralized, two-agent swarm robotic system that integrates a terrain-adaptive rover with collaborative navigation and mapping capabilities. A rocker suspension-based rover optimized for rugged environments was prototyped and localization was achieved through encoder-based odometry, with swarm coordination executed using finite state machines and ESP-NOW peer-to-peer wireless communication. Simultaneously, multi-agent SLAM and decentralized behavior were validated in ROS2 and Gazebo using RTAB-map. A monocular depth estimation pipeline based on transformer architectures was implemented in simulation to predict depth from lunar imagery. Results from hardware trials and simulation-based experiments demonstrate the feasibility and modularity of the proposed architecture, forming a scalable foundation for future swarm missions.