Multi-robot systems offer significant advantages over single-robot approaches in terms of adaptability, fault tolerance, and efficiency. However, a primary challenge in deploying low-cost robot swarms is maintaining accurate localization over time, as reliance on low-cost proprioceptive sensors (wheel encoders and IMUs) leads to accumulated odometry drift. Equipping every agent with high-precision environmental sensors and sufficient computing power for Simultaneous Localization and Mapping (SLAM) effectively solves this issue but negates the cost effectiveness and scalability of the swarm. This thesis presents the design, implementation, and evaluation of a cooperative localization framework within the Robot Operating System 2 environment. The system uses a heterogeneous "Leader Follower" architecture to address trade-off between localization accuracy and system cost. The architecture designates a single, computationally capable "Leader" robot equipped with a 2D LiDAR to function as a mobile perception hub, while "Follower" robots rely on minimal onboard sensing. The methodology integrates three core technical components: Global Perception, Relative Tracking, and Sensor Fusion. The framework was validated using a high-fidelity physics simulation in Gazebo. Real life hardware framework has been developed. Experimental results compare the trajectory accuracy of Followers relying solely on dead reckoning against those corrected by the cooperative system. The analysis demonstrates that the implemented framework reduces localization error, effectively mitigating sensor drift and enabling stable formation control without requiring expensive sensors on individual Follower units. This work confirms the viability of LiDAR based relative tracking for low-cost swarm coordination in GPS-denied environments.