Due to the high packet loss rate during multi-hop transmissions in wireless sensor networks, a reliable end-to-end delivery scheme is desired. But, in some applications, there is no need of 100% communication reliability (CR). Active caching is a novel scheme which can satisfy the flexible CR. However, in this scheme, only caching nodes can detect missing packets, so the retransmission efficiency is low. In this paper, we propose a more efficient scheme named OAC (optimized active caching) which exploits the broadcasting nature of wireless communication to detect packet loss. The results based on simulation show that the proposed scheme has advantages in retransmission counts and time-delay. And thus energy is saved at the same time.Keywords-reliable transmission; packet loss detection; optimization; broadcasting nature; wireless sensor networks I. NTRODUCTION Wireless sensor networks (WSNs) could be comprised of hundreds of thousands of sensor nodes which consist of sensing, data processing and communicating components [1]. Because of the recent technological advances in micro electro-mechanical systems [2], WSNs have attracted extensive attention of business world and academic circles during the last ten years. There have been a lot of applications such as environment monitoring [3], localization system [4], home automation, automated assistance for the elderly [5] and so on. Within these applications, sensor nodes collect data of interested events from the surroundings and deliver them to the sink node which can communicate with the external network.Due to the high packet loss rate during multi-hop transmissions in wireless sensor networks, we need a reliable recovery mechanism to retransmit the lost packets between the sensor nodes and the sink. In conventional networks like the Internet, we have many transport layer protocols for use such as the TCP protocol. Although these protocols can provide the reliable end-to-end transmission, congestion and flow control, it is unsuited for sensor networks because of the specific requirements (such as high energy efficiency, low computation complexity of algorithms, data-centric and so on) and the extreme communication conditions that sensor networks exhibit [6]. Therefore, a reliable recovery method for WSNs is desired.The applications of WSNs are application-dependent and have different emphases on metrics. For instance, within monitoring applications, the reliability is the most important; for event detection, timeliness is crucial [7]. So, there is no need to consider 100% reliability in some applications. For example, in multimedia sensor networks, the traffic usually has high bandwidth demand and strict delay bounds. So, video streams can be within a certain level of distortion [5]. Besides, as we all know that, energy efficiency is the most important concern in the design of WSNs. A permissible level of loss tolerance can reduce the requirement of memory, lower the transmission numbers, and thus save the rare energy [8].So, in order to achieve various levels ...