This paper presents an agent-based algorithm for exploring enclosed and uncharted environments, with the primary objective of locating scattered targets. The algorithm addresses key limitations of swarm agents, including restricted perception, limited communication, and constrained memory. It incorporates four exploration strategies – inertia action, virtual attractive force, virtual repulsive force, and random selection – allowing agents to adapt by randomly selecting one at each iteration. The algorithm is evaluated in two simulated scenarios featuring varying agent types and environmental obstacles. Results demonstrate its effectiveness: the swarm consistently achieves target localization while covering approximately 80% of the environment in fewer than 5000 iterations. These findings highlight the algorithm’s capability to enable efficient exploration through decentralized decision-making under partial information. The proposed approach offers a promising alternative for autonomous exploration tasks involving multi-agent systems in constrained or dynamic environments.