Communication

In our NCS benchmarking platform, we have two entities, i.e., the Lego Mindstorm Robot and the Controller, communicating through a network. This happens via exchange of sensor and actuation messages through UDP sockets.

Robot-to-Controller

The Robot periodically samples gyro rate, left and right motor position values with a few selected timestamp values. Subsequently, it sends them through the socket. Once the Robot loop starts, measured sensory information is packed into a structure of 10 'long long' defined in src/packet.py:

R2H_PACKET_FORMAT = 'qqqqqqqqqqq'
R2H_PACKET_SIZE = struct.calcsize(R2H_PACKET_FORMAT)

One can add / remove more data fields to transfer more / less data based on the requirements.

Controller-to-Robot

Based on the latest measurements, the controller replies to the robot with a set of actuation signals:

H2R_PACKET_FORMAT = 'qq%sq' % (NO_OF_DELAYS * NO_OF_VOLTAGES_PER_DELAY)
H2R_PACKET_SIZE = struct.calcsize(H2R_PACKET_FORMAT)

The actuation signals are voltage values to apply to the left and right motors for a given set of sampling periods. This is done for delay compensation in case of lost or delayed information. The size of the H2R_PACKET_FORMAT variable is dependent on the selected control policy for delay compensation. It can be tuned by changing the control policy and the variables NO_OF_DELAYS, NO_OF_VOLTAGES_PER_DELAY accordingly.

For more information please refer to our paper.