Speaker: Milos Rovcanin
Date: 15 October 2013
This demonstration illustrates how independent networks can work together to improve their performance in terms of throughput, reliability, delay and reduced energy consumption. In the demonstration, two independent sensor networks are set-up for different purposes: one for monitoring the temperature of a building and a second one for security monitoring. Both networks are capable of cooperating with each other by dynamically activating and configuring network optimization techniques based on their network requirements. The networks are capable of learning about the optimal network settings without any required a-priori configuration. A reinforcement learning agent is used to continuously monitor the network performance and to change the configuration when network conditions or application requirements change. The result is an improved network performance that learns the best reaction to be adaptive to changing network conditions.