In this paper, we have introduced the design, implementation
and evaluation of multi sensor data retrieval strategies for
cloud robotic systems. We proposed an architecture consists of
a data center, cloud cluster hosts and robot clients. In addition,
we tackled the problem of MSDR among the host-based
framework by defining the problem into a Stackelberg game and
offered theoretical optimization analysis. Our proposed scheduling
scheme with a data buffer are implemented in the cloud
cluster host module. In order to evaluate the proposed strategies,
we define the QoS criteria that is used in the experiments. Our
experimental results demonstrate significant improvement of
the proposed approach in terms of ToR, RoR, bandwidth usage,
and CPU load, by adopting the proposed strategies for resource
retrieval. For future study, aiming at the optimization of data requirement
for dynamic robotic tasks, the scheduler will be explored
concerning a prediction model for the completion time
of the required data.