我們在本篇論文提出一個透過多躍距收集資料的路由協定—BHAR(Bidirectional Hierarchy-based Anycast Routing)，和HAR做比較，BHAR加快了建構階層式路由樹及修復路由的速度，並改善其運作機制。在BHAR中，匯集節點及來源節點皆能建立階層式路由樹，各個節點僅需依靠對其自身的親節點及鄰居節點的認識加入路由樹；來源節點所建立的路由樹可藉由與其它相鄰路由樹的路由交換，學到至遠端匯集節點的路由，並透過中繼路由樹節點將資料送達最近的匯集節點；在不依靠位置資訊、不需遠端控制的情況下執行路由修復，同時做局部網路的最佳化調整，以避免定期及條件式重建網路拓樸對網路及資料傳輸所造成的衝擊，並改善HAR在路由修復上修復彈性較差及仍有可能發生迴圈的缺點。我們使用VC++對BHAR與HAR做效能比較，從模擬結果來看，BHAR在匯集節點與來源節點數目增加時，可減少一般節點加入路由樹的平均等待時間；總節點數及網路大小的增加對BHAR的效能衝擊也相對小很多；BHAR的可擴充性(scalable)也比HAR大許多；在修復路由的能力上，BHAR比HAR能容忍更多的節點故障，並維持有效的路由運作，大幅提高了網路的強健性(Robustness)；在節點因各種因素導致分佈不平均的情況下，BHAR在路由修復的能力上比HAR優秀甚多。 In this paper, we present Bidirectional Hierarchy-based Anycast Routing (BHAR), a routing protocol for collecting data over multi-hop. In comparative to HAR, BHAR speeds up and improve the mechanism to construct hierarchical trees and to repair the route. In BHAR, sinks and sources can initial to construct a hierarchical tree. By knowing only its own parent and neighbor, each node in BHAR can join a tree; a tree can learn the routes to a remote sink by exchanging its routes with its neighboring trees, and send data to the nearest sink by intermediate trees; each node can perform route repair without geographical information or being controlled remotely, and perform local network topology optimization simultaneously in order to prevent the impact to the network and to data communication by periodically and conditionally network reconstruction. We evaluate the performance of BHAR by using VC++ and comparing with those of HAR. The simulation results demonstrate that the average waiting time for a node to join a tree decrease when the number of sinks and sources increase in BHAR. The impact of the increase of the node number and network size to the performance of BHAR is relatively small, which means the scalability is much better. BHAR can endure more nodes to be failed and maintain effective routing operation, which has improve the robustness of BHAR. BHAR achieves much higher performance on route repair in the situation of the unevenly distribution due to different factors.