We introduce a novel method for latency compensation in realtime critical cloud or edge control scenarios for mobile robots. Offloading computation towards virtualized computing resources provides several advantages along the software life cycle and facilitates scaling and extensions of the respective software services or stacks. In former work, we introduced the modularization and offloading of even realtime critical control functions for autonomous navigation to an edge computing infrastructure through the usage of low-latency 5G networks. In order to overcome performance or operation reductions when facing latency peaks or temporary connection losses, we propose a predictive control method that generates control hypotheses, i.e. sampling, even for future steps. On the client device, the hypotheses are matched with current sensor readings if control commands fail to appear. In this publication, we focus on the efficient hypotheses matching on the client side and identify an adequate hashing approach for planar laser scanner data. Within different simulation scenarios, we show the general feasibility and benefits of our approach.