Test bench. 3.1. The test bench shall be constructed as follows:
Test bench. As a proof of concept, we implemented a small test bench in the present package cbg-executor_ping- pong_cpp. The test bench comprises a Ping node and a Pong node which exchange real-time and best-effort messages simultaneously with each other. Each class of messages is handled with a ded- icated Executor, as illustrated in the following figure. we validated the functioning of the approach. With the test bench, In this example, the callback for the high priority task (red line) consumes 10ms and the low priority task (blue line) 40ms in the Pong Node. With a ping rate of 20 Hz, the CPU saturates (10ms*20+40ms*20=1000ms). With higher frequencies the high priorty task can continue to send its pong message, while the low priority pong task degrades. With a frequency of 100Hz the high priority task requires 100% CPU utilization. With higher ping rates it keeps sending pong messages with 100Hz, while the low priority task does not get any CPU ressources any more and cannot send any messages. The test bench is provided in the cbg_executor_demo.
Test bench. As a proof of concept, we implemented a small test bench in the present package cbg-executor_ping- pong_cpp. The test bench comprises a Ping node and a Pong node which exchange real-time and best-effort messages simultaneously with each other. Each class of messages is handled with a ded- icated Executor, as illustrated in the following figure. With the test bench, we validated the functioning of the approach - here on ROS 2 v0.5.1 with the Fast-RTPS DDS imple- mentation - on a typical laptop. The test bench is provided in the bg-executor_ping-pong_cpp package of the micro-ROS_experiments repository.
