
Chapter 5.
Evaluation
This chapter deals with the evaluation of different parts of the system. Two test scenarios are
described, one office and one home environment as well as various simulations to test the system’s
performance.
5.1. Benchmarking
The individual components of the software were tested in terms of their latency, processing time
and memory usage. All tests were performed on an Asus UL30VT-QX061V laptop with an Intel
U7300 processor with 2 x 1.3 GHz and 4 GB Memory, running the Ubuntu 12.04 LTS operating
system. Figure 5.1 shows the structure of the test system.
gateway driver
devicemanager
device
operations
visualization &
namespace
simulation
node
protocol
ns/device
operations
Figure 5.1.: The structure of the software chain for testing the performance of the different nodes.
An additional ROS node was implemented to inject artificial data into the different topics, as
well as calling all services with any speed or repetition number. This simulation node allows
custom data structures to testing special cases and the robustness of the system with regard
to faulty data. However, the actual performance tests are done within the three main nodes
and not within the simulation node in order to achieve accurate measurements. Otherwise, the
processing and communication time of the simulation node would distort the benchmark results.
All times are measured with Qt’s QTime class, which offers functions for absolute and relative
time measurements with 1 ms resolution.
65
Comentarios a estos manuales