This paper studies automated bilateral negotiation among self-interested agents in complex application domains which consist of multiple issues and real-time constraints and where the agents have no prior knowledge about their opponents' preferences and strategies. We describe a novel negotiation approach called OMAC (standing for “Opponent Modeling and Adaptive Concession”) which combines efficient opponent modeling and adaptive concession making. Opponent modeling is achieved through standard wavelet decomposition and cubic smoothing spline, and concession adaptivity is achieved through dynamically setting the concession rate on the basis of the expected utilities of forthcoming counter-offers. Experimental results are presented which demonstrate the effectiveness of our approach in both discounting and non-discounting domains. Specifically, the results show that our approach performs better than the five top agents from the 2011 Automated Negotiation Agents Competition (ANAC).