Several months ago, I posted a comment on Greater Greater Washington in which I discussed certain aspects of the design of the Metrorail system, including the use of bi-directional signalling, and the spacing of interlockings along the mainline. Given that bi-directional signalling is one of the more esoteric aspects of rail signalling design, I figured I would write a post on the value of bi-directional signalling and some of the technical aspects involved.

I'm not going to provide a whole primer on fixed-block signalling here (maybe in a later article), but I can endorse this guide to signalling on the New York City subway and this page on mainline rail signalling in the UK.

In its most basic configuration, a fixed-block signalling system is designed to protect trains moving in a specified direction on each track. This is reflected in the design of the signalling system—signal heads face trains moving in that direction, track circuits and interlockings are configured accordingly, and so on. This holds true even for systems which use cab signalling rather than conventional wayside signal heads. When trains move against the direction of traffic, they receive no protection whatsoever from the signalling system. Accordingly, wrong-direction moves can be performed only under strict control by supervisory staff (on the London Underground, for example, it requires filling out paperwork).

Bi-directional signalling changes that. With bi-directional signalling, the signalling system is able to protect trains moving in either direction. Doing so, though, requires more equipment. The interlocking—whether electromechanical, relay-based, or solid-state—requires more logic, which increases the complexity (and thus cost). In addition, signal heads must be provided facing in both directions, along with stop arms, on systems that use them. (Note that there are some exceptions; on PATH, for example, not all reverse-direction signals have trainstops). A more extensive discussion of the infrastructure required for bi-directional signalling is given in this article on nycsubway.org.
On many older systems, bi-directional signalling isn't the norm. As a result, these systems have few choices when they must take a track out of service. Without signal protection, running trains against the direction of traffic is generally done only in non-revenue service. This is in stark contrast to operations on Metrorail, where single-tracking done routinely when a track must be taken out of service. For some older systems, new lines, line extensions, and resignalling projects provide an opportunity to add the necessary facilities for bi-directional operation. Consider this account from Leonardo, et al. on signalling for the (relatively new) Sheppard Line in Toronto:

The majority of the original TTC subway is signalled for single direction travel only. Reverse traffic signalling was only provided in very limited areas where a specific operational need was identified. On the Sheppard Line, a limited reverse traffic capability has been provided. This capability allows for single track operation between interlockings in the event of a blockage on one track. A total of 13 additional signals were provided in order to achieve a limited reverse traffic capability, which safely supports signalled operation in the reverse traffic at low speeds. The achievable headway is constrained by the relatively wide spacing between crossovers and the low operating speeds, but the facility has already proven useful in the first six months of operation. It is expected that greater use of the feature will occur in future years as operating confidence grows, and as maintenance activities inevitably increase with the age of the infrastructure.

One of the great advantages of moving-block signalling systems, like Thales (formerly Alcatel) SelTrac or Siemens (formerly MATRA) Trainguard MT is that bi-directional operation is ordinarily built-in. This is used to great advantage on systems like the DLR and AirTrain JFK. However, I was surprised to find out that the ongoing SelTrac deployment on the London Underground Jubilee Line will not support bi-directional operation, which I consider rather short-sighted.

So, what's the good in having bi-directional signalling? As noted above, bi-directional signalling provides substantial additional flexibility in off-nominal conditions, by providing trains with protection when they operate against the normal direction of traffic. However, there's an additional benefit for systems with more than two tracks—with bi-directional signalling, one or more of those tracks can be used as reversible express tracks. This is the situation on the Flushing Line in New York City, for example, where the center track is used by inbound express trains during the morning rush hour, and outbound express trains during the evening rush hour. At other times, the center track provides additional operational flexibility for getting around construction or disabled trains.

I often see comparisons made to the New York City subway in which it is argued that four tracks are better than two, not merely for enabling express trains, but also for making it possible to keep trains running when tracks are taken out of service or trains become disabled on the mainline. However, I would argue that a two-track system can do quite well, so long as the signalling system makes it possible to keep trains running during disruptions.