The concept of in situ resource utilisation (ISRU) was proposed by the USA National Aeronautics and Space Administration (NASA) to extract off-Earth resources from the Moon, asteroids, and Mars to maintain successful long-term manned space missions. The focus of this paper is to develop an optimal swarm robotics system in terms of its design and operation strategy, and to consider its application for extracting water from the super frozen craters at the Moon’s poles. The optimisation of swarm robotics system has been conducted in terms of fleet size, collision avoidance and convergence time. Analytical research, numerical modelling and simulation have been applied to design and evaluate the performance of swarm robot fleets for off-Earth mining. This involved determining the optimal number and dynamic interaction of robots in the fleet to achieve the best swarm performance for the considered task. The proposed swarm robotics system has been designed based on the behaviour and autonomy of swarm insects. Its performance has been evaluated using swarm intelligence (SI) classification, ie natural and artificial swarm intelligence using numerical simulations. To study the swarm robotic system design, the behaviour of two different insect colonies, namely the ant colony and the bee colony, has been studied. After comparative evaluation of these methods, the artificial bee colony (ABC) optimisation method has been used to construct a swarm robotics system, as it has been deemed more suitable for the considered task because it does not require the feature for pheromone tracking. Based on the collision and signal error analysis for different size swarm robot fleets, the characteristics of swarm robotics systems (eg scalability, robustness, and flexibility) have been evaluated. Assuming that the maximum robot signal coverage is 300 m, and the maximum movement speed of the robot is 150 m/h, the swarm fleet size of 10 robots with 0 per cent signal and collision error has been identified as the most optimal. The convergence time for 10 robots is 150 seconds, and for 15 and 20 robots is 200 seconds and 275 seconds, respectively. The findings presented in this paper can also be utilised and beneficial for on-Earth mining and other geotechnical engineering activities.