CSOnet’s architecture was designed to be a set of local WSAN’s that connect to an existing wide area network (WAN) through gateway devices. CSOnet can therefore be viewed as a heterogeneous sensor-actuator network. It consists of four types of devices:
The following figure shows the prototype CSOnet built by Ruggaber et al [1]. This network shows a single ANode (marked by the ”V”) that receives its feedback sensor signals from three INodes (marked by the ”I”). One of the INode’s is located at the river to monitor actual CSO discharge into the river. The other INode’s are used to measure the water level in a retention basin. The distance between the CSO outfall at the river and the retention basin is about 3 miles. Feedback information from the CSO outfall is forwarded over a a line of RNodes (marked by the ”R”). This figure can be taken as a very simple example of a singleWSAN. This particularWSAN has been in continuous operation since 2005 and has been extremely useful in refining CSOnet’s hardware and middleware components to ensure long-life and economical maintenance.
In order to scale CSOnet up to an entire metropolitan area, it was necessary to adopt a hierarchical architecture consisting of several WSAN’s that are interconnected over an existing wide area network. One reason for this is the well-known inability of WSAN’s to provide an acceptable quality of service when the network becomes too large. This is a consequence of the wellknown theoretical limitations on wireless network throughput [2]. Empirical studies from DARPA’s NEST program [3] have suggested that flat WSANs should be limited to a diameter of 5-6 hops to prevent excessive congestion. If South Bend’s CSOnet were to be built as a single flat WSAN covering the entire city, it would consist of several hundred nodes covering a 13,000 acre area. Such a deployment would require a prohibitive number of INodes and RNodes. For this reason, CSOnet consists of a set rather small WSAN’s that forward their data to GNodes. The GNodes then forward the received packets to other WSAN’s in the system.
To understand CSOnet’s hierarchical structure, we first need to examine the actual sewer system to be controlled. The following figure shows a sewer system in which combined sewer trunk lines feed into a large interceptor sewer. Prior to 1972, municipal combined sewer lines were allowed to dump directly into rivers and streams. Under the clean water act, cities were forced to treat the water from these combined sewer lines before they were released into a river or stream. One common way to meet this regulatory burden was to build a interceptor sewer along the river. This sewer would intercept the flow from the combined sewer trunk lines and convey that flow to a wastewater treatment plant (WWTP). Under dry weather conditions the flows were small enough to be handled by the WWTP. Under wet weather conditions (storms), the flows often overwhelmed the WWTP’s capacity, thereby forcing operators to dump the excess directly into the water. As noted above such discharges constitute the CSO events described earlier.
From the above figure we can see that the combined sewer trunk lines and interceptor sewer connect at a CSO diversion structure. This is the point where we can apply control. The current system in South Bend Indiana uses a passive thresholding control. When the depth of the flow is below a fixed preset threshold, the flow is diverted into the interceptor sewer. Above this threshold, the flow is dumped out into the river. The problem is that these thresholds are set for the worst-case storm scenario. By placing a WSAN in the combined sewer trunk line above the CSO diversion structure, we can estimate the actual flows into the interceptor line and thereby make closed-loop control decisions that optimize the flow into the interceptor line such that WWTP capacity limits are never exceeded. This means that the natural place to place ANodes is at the CSO diversion points. These ANodes would then adjust the amount of water diverted into the interceptor sewer line based on an adaptive threshold that is a function of the current flows into the system. Because this scheme is adaptive, it need not be as conservative as the original passive thresholding scheme.
The scenario outlined above therefore indicates that CSOnet
consists of a collection of WSAN’s that forward flow measurements
in a combined sewer trunk line to its associated
CSO diversion structure. At this diversion structure would be
a GNode and ANode. The ANode would adjust the flow into
the interceptor line and the GNode would serve as a gateway
between this particular WSAN and neighboring WSAN’s up and down the interceptor line.
The following figure illustrates this system
architecture with 2 different WSAN’s controlling the two
diversion structures into the interceptor line. GNodes at these
diversion structures and the WWTP are used to exchange
control information in a way that allows coordinated flow
control across the city’s entire sewer system.