This paper presents Random Motion Differential Evolution (RanMoDE) as a method for online indoor localization of flight robots using a minimal sensor setup. The sensor setup consists of six directional 1D infrared distance sensors and an inertial measurement unit (IMU) with an attitude heading reference system (AHRS). Pose estimation is performed offboard within a digital environment model, without relying on explicit motion models or parameterized probabilistic sensor likelihoods. RanMoDE formulates localization as population-based online optimization, combining randomized motion for exploration with differential evolution for exploitation, and evaluates pose hypotheses solely by comparing real distance measurements with ideal raycast-based virtual measurements. A systematic parameter study in a virtual industrial machine hall identifies a standard configuration that enables stable operation at an update rate of 50 Hz. Simulation results demonstrate robust online localization even under strong measurement noise, achieving a success rate of 100% up to a sensor error of emax = 0.6m. Real flight experiments confirm the practical feasibility, achieving a position root mean square error (RMSE) of RMSEpos = 0.125m and a rotation error of RMSEψ = 0.049 rad at average computation and transmission times of 15.4 ms. Short-term model deviations and initialization errors are reliably compensated. RanMoDE combines low sensor and integration effort with high robustness, making it particularly suitable for flexible and energy-efficient flight robot applications in reconfigurable production environments.