The mobile Ad Hoc networks composed of mobile terminal nodes with wireless communication device are typically characterized by their multi-hop,temporary and un-center.It can be quickly built up in anywhere and at anytime and every terminal can move freely and be equal.The topology of Ad Hoc is dynamically changing.The bandwidth and energy are also limited,etc. Because of these, the research on the Ad Hoc network, in particular, the research on its routing protocols is immensely challenging.Ant colony algorithm is a kind of swarm intelligence heuristic approach that inspired from social insects in nature. Swarm intelligence refers that one agent can only do some easy jobs, and very complicate tasks must accomplished by the cooperation of each agent in a whole colony. A typical example of swarm intelligence is ants find food. Ant colony algorithms have widely applied to solve many combinatory problems. Ant colony algorithms support distributed computing and multi-path, the agents are also easy to implement. So ant colony algorithms just meet the need of mobile Ad Hoc network routing.Routing protocol is one of the key technologies in Ad Hoc which plays a very important role in its network performance. This paper carries out in-depth analysis and research in the design principles and ideas of Ad Hoc routing protocols. It also introduces researches in this field so far, suggests the classification of Ad Hoc network routing protocol. Three typical routing protocols(AODV, DSDV DSR) were simulated and analysed by NS2 simulation tool. The advantages of AODV were proved and its insufficiencies were also pointed out. Then this paper presents the basic rules and ways of ant colony algorithm and discusses the feasibility to apply it to Ad Hoc networks.Based on the above research, Combining advantages of both ant colony algorithm and Ad Hoc network, This paper proposes an Ad Hoc networks routing algorithm based on Ant Colony(ANRBA) through extending the AODV routing protocol. The Algorithm improves the formation and maintenance of AODV routing strategy, it has the ability of distributed parallel processing, a lot of redundant paths is provided, and the survivability of network is enhanced. At the same time, the algorithm adaptively adjusts the probability routing table by using the networks information of the ants collected and achieves globally optimizing, it provides a new method to improve the congestion problem and heavy overheads problem. The simulation results show that the ANRBA performed well in packet delivery rate ,the average end-to-end packet transmission latency, normalized routing overhead and it achieved an improving performance.