This paper presents EARA-CG, an improved version of a swarm-intelligence ad hoc routing algorithm EARA introduced in [10]. Based on the understanding of the evolutionary cooperation in the biological swarm, we use the principle of swarm intelligence to reinforce good quality routes with only local communication. The concept of stigmergy is adopted to reduce the amount of control traffic. By locally monitoring the network transmission queue and other MAC layer information, this algorithm can forward data traffic through paths that avoid network hotspots. The data traffic influences routing table updates at each node. We also discuss an evaluation methodology to simulate ad hoc networks, and the simulation results show that this novel routing algorithm scales well to a variety of network conditions.