In this paper, we investigate the problem of multi-user resource management in multi-hop cognitive radio networks for delay-sensitive applications. Since the tolerable delay does not allow propagating global information back and forth throughout the multi-hop network to a centralized decision maker, the source nodes and relays need to adapt their actions (transmission frequency channel and route selections) in a distributed manner, based on local network information. We propose a distributed resource management algorithm that allows network nodes to exchange information and that explicitly considers the delays and cost of exchanging the network information over the multi-hop cognitive radio networks. The term "cognitive" refers in our paper to both the capability of the network nodes to achieve large spectral efficiencies by dynamically exploiting available frequency channels as well as their ability to learn the "environment" (the actions of interfering nodes) based on the designed information exchange.Note that the node competition is due to the mutual interference of neighboring nodes using the same frequency channel. Based on this, we adopt a multi-agent learning approach, adaptive fictitious play, which uses the available interference information. We also discuss the tradeoff between the cost of the required information exchange and the learning efficiency. The results show that our distributed resource management approach improves the PSNR of multiple video streams by more than 3dB as opposed to the state-of-the-art dynamic frequency channel/route selection approaches without learning capability, when the network resources are limited.Index Terms: distributed resource management, cognitive radio networks, multi-hop wireless networks, multi-agent learning, delay sensitive applications.