Wireless sensor networks (WSNs) are task-specific where the tasks are specified by the users or applications. Energy and delay constraints have had a great impact on the design and operation of the WSNs. As routing plays an important role in data-centric WSNs, there are numerous energy aware routing protocols proposed for WSNs with single sink, generally using multi-hop paths to use the energy more efficiently. But most of them do not consider the communication delay carefully. As many applications of WSNs require real-time communication, the end-to-end delivery delay is another point to be considered, especially for urgent events. When considering the multiple QoS constraints such as the delivery delay and the minimal energy required, the multi-constrained routing is NP-hard.

Moreover, WSNs also suffer from a scalability problem as the total number of nodes can reach hundreds or thousands. Thus, sensors located near the sink consume large amounts of energy. The sink is then isolated once they deplete their batteries, and the network fails to function. To cope with the scalability problem, multi-sink configuration is adopted for large-scale WSNs. How to select the optimum sink for packets with different delay and energy constraints in large-scale WSNs is the main problem to be resolved in this paper, which is called the multi-constrained anycast (MCA) problem. Obviously, it is NP-hard.

We investigate the MCA problem in resource-constrained WSNs with multiple sinks and two-tier architecture, where the source cluster-heads send collected data to only one optimum sink. To reduce the energy consumption and shorten the delivery delay, a distributed anycast algorithm ARMCA inspired by swarm intelligence of ants is proposed, which can also achieve reliable data gathering during the deployment period. The scheme takes advantage of beneficial clustering and routing emerging in a self-organized way from the positive interaction of ants. The idea of the ACO-based routing is adopted to achieve the good performance of energy-efficiency and delivery delay. Meanwhile, the ant-based clustering is utilized to ensure the reliability. By the combination of the ant-based routing and clustering models, the ARMCA protocol can solve the MCA problem efficiently and flexibly.

The simulation results demonstrated that the proposed algorithm can reliably deliver collected events to the optimum sinks efficiently, and has self-recovery capabilities to deal with failures of sensors including the sinks. Not only is the energy cost reduced, but also the delivery delay is shortened significantly when transmitting packets of the named events to the sinks.