This paper investigates the problem of moving a mixture of active and passive elements to a desired location using a swarm of wheeled robots that require only two bits of sensory input. It examines memory-less control strategies that map a robot’s sensory input to the respective wheel velocities. Results from embodied simulations show that the problem can be solved without robots having (i) to discriminate between active and passive elements or (ii) sense other robots. Strategies optimized for moving passive elements, or mixtures of active and passive elements, performed robustly when changing the mixture of elements, or scaling up the number of robots (up to 25) or elements (up to 100). All strategies demonstrated to be fairly robust to noise and adaptable to active elements of different dynamics. Given the simplicity of the robot capabilities and strategies, our findings could be relevant in scenarios where microscopic swarm robots need to manipulate mixtures of elements of unknown dynamics, with potential applications in nanomedicine.