TRAFCOD: A Method for Stream-Based Control of Actuated Traffic Signals

In the TRAFCOD method, control of traffic signals at an intersection is formulated in terms of traffic streams, without reference to stages (sets of streams that are green together). The resulting formulation is more flexible and direct than stage control, allowing for easy implementation of a variety of traffic control tactics, including priority tactics for public transport, demand responding realization and coordination between streams. Control is expected to follow a given cyclic sequence structure; however, streams with no demand may be skipped, allowing the control program under some circumstances to uncouple into rings that cycle independently. Sequence state is formulated using an array of successor functions called the sequence array (for each conflicting stream pair, the array element is either 0 or 1 depending on which stream has the next turn). Changes in the sequence state are effected by rotations of the sequence array. Control formulas are expressed in terms of standard traffic engineering inputs, allowing for automatic generation of control formulas. Related TRAFCOD software includes a program for automatic formula generation and a keyboard-actuated controller simulator used for training traffic engineering students. Tests confirm that the model works as expected. Vehicle actuated intersection control means that the timing and, in more advanced controllers, the sequence of the signal indications are responsive to vehicle detections. While timing decisions are important, the focus of this paper is sequence control; in other words, if several conflicting streams have requests, which will become green next? Flexibility in sequence control depends on the controller hardware and the software that drives the controller. Both controller hardware and software can be classified as either stream-based or stage-based. (A stream is a traffic stream with its own signal indication; a stage is a set of streams that are green simultaneously. More complete definitions follow in the body of the paper.) Stream-based hardware, which is standard in most of northern Europe including Great Britain (1), allows most of the control logic to be formulated in user-programmable software that outputs to the controller the signal indications. Compared to the traditional stage-based controller, stream-based controllers offer advantages for both actuated and fixed time control (2, 3). From a traffic engineering point of view, there are three main arguments for stream-based control. First, efficiency – clearance times can be enforced stream to stream rather than stage to stage, leading to less lost time. Second, simplicity – stages and overlaps need not be spelled out, but simply occur as a byproduct of how the streams follow each other. Third, flexibility – because most traffic control tactics relate directly to streams rather than to stages, stream-based control more easily allows the traffic engineer to apply desired control tactics. For example, minimum green time is meaningful for streams, but not for stages. Likewise, most tactics for safety, coordination, and priority are most easily formulated at the stream level. For example, with stream-based control, phase insertion becomes simply a matter of checking for conflicting streams, without the need to specify all the stages to which a priority stream could possibly belong. During periods of light traffic, stream-based control makes it easy to apply tactics for safety and efficiency such as skipping no-demand streams and waiting in red. The application of various safety tactics is often one of the main reasons for applying stream-based control in the Netherlands and in Scandinavia (4). Stream-based controllers have been called acyclic (5) because they don’t require a cyclic service sequence, although they can be used with a cyclic sequence as well. Of course, because a stage is a set of streams, the software driving a stream-based controller can still be formulated in terms of stages. Surprisingly, stage-based control appears to be the norm with vehicle-actuated signals not only in the United States, where stage-based hardware dominates, but also in some countries with stream-based controllers such as Germany (6) and Great Britain (including the MOVA control system) (7). While American double-ring controllers (8) and German control software allow some responsiveness in stage sequence, they are still less flexible than stream-based control. The stream-based control language TRAFCOL (TRAFfic COntrol Language) and derivative products have been used in the Netherlands since the early 1970’s. An abbreviated version (omitting administrative functions such as checking that detectors are working) called TRAFCOD (TRAFfic COntrol Design) (9) was developed in 1978 to train traffic engineering students. However, even though control languages used in the Netherlands are stream-based, their structure and / or application often place artificial restrictions on stream sequence logic. For example, one common restriction is to require that a fixed stage appear every cycle to provide a clear cycle start, regardless of demand, in order to simplify sequencing logic. Another popular software product (10) controls sequence by dividing the cycle sequence into modules, which are a kind of super-stage, with rules for allowing the program under various circumstances to cross module boundaries. While the rules that allow for such exceptions restore a good deal of flexibility, they are often not used to full advantage, depending on the expertise of the traffic engineer. Just as important, the ad hoc nature of the sequence logic in existing control methods usually requires that detailed specification of control formulas which rely on specialized knowledge and are prone to error. Research on stream-based sequence control has recently led to a method, now incorporated in an updated version of the TRAFCOD language, that appears superior in several respects to other known sequence control methods. It is based on a very general model of sequence state that is extremely flexible, with no reference to stages, modules, or streams that must have green each cycle. With this model, most traffic control strategies, including various priority tactics for public transport, can be formulated in terms of basic traffic engineering data, allowing control formulas to be generated automatically. Instead of spending a lot of effort to make sure that control formulas are cleverly and correctly specified, the traffic engineer can instead concentrate on higher level decisions such as which traffic control tactics to use. The following section clarifies the terminology of stream-based control. Next the new model of stream sequence used in TRAFCOD is explained. The main section of the paper lays out the principal TRAFCOD sequence control formulas, followed by some illustrative examples. The last sections describe additional control tactics and the status of ongoing work with TRAFCOD. Streams, Stages, Phases, and Activation A traffic stream or stream is the smallest unit of control in the intersection. Generally, it is the set of movements sharing the same queue and controlled by either a single signal or a set of signals that must always give the same indication. Two streams are conflicting if they may not have the right of way simultaneously. A clearance time matrix expresses the minimum required interval between the start of stream i’s red and stream j’s green for each conflicting i,j pair. Clearance times depend on intersection geometry and stream speeds as well as local policy, and are therefore not generally symmetric. Various types of coordination constraints may apply to non-conflicting stream pairs. Simultaneous start means that a pair of streams must start green at the same time, unless one is skipped for lack of demand. This tactic is mostly applied for safety reasons to opposing traffic streams with permitted left turns, in order to help establish priority for the through movements. It may be implemented with or without a common request, meaning that a request on one stream is taken as a request on the other. There are also various types of staggered start constraints, in which one stream’s green is to start a certain interval after another’s. Staggered start can be used for safety reasons to help establish priority (e.g., have a bicycle or pedestrian stream begin 2 s before a parallel general traffic stream with permitted right turns). It can also be used to provide a green wave for movements through two stop-lines such as pedestrians using a series of crosswalks, or vehicles at complex intersections. A stage is a set of streams that receive the right-of-way simultaneously. Viewed globally, an intersection’s signals can be seen as going from one stage to another. Elsewhere, stages are sometimes called phases. However, we use the term phase to refer to the state of a stream’s signal, the main phases thus being green, yellow, and red. In the Netherlands, the green phase is generally divided into five subphases, arranged serially as follows: * advance green (GA) – used only with certain staggered start tactics, e.g., to hold a stream in green until a coordinated stream becomes green. * fixed time green (GF) – a green time of fixed duration. For actuated control, this sub-phase is long enough to let the discharge flow stabilize (say, 6 s). For fixed time control, it is the entire green phase. * waiting green (GW) – the sub-phase in which a stream waits if there is no request on a conflicting stream (this sub-phase is skipped for some tactics such as waitin-red). * extension green (GX) – green time granted on the basis of extension requests; it ends when there is no more extension request, or after a specified maximum duration. * parallel green (GP) – additional green time granted because no other stream can take advantage of the subject stream’s green ending, usually because a parallel stream is still in one of the first four green sub-phases, preventing waiting streams from becoming green. When a stream becomes active, it seizes control of the program, forcing conflicting streams to deactivate, allowing the subject stream to become green, and preventing conflicting streams from being activated. The active period includes first four green sub-phases and a short red period preceding the green start during which deactivated streams have their yellow and clearance times. Model of Cyclic Stream-Based Sequence Control Sequence Structure With cyclic sequencing, the traffic engineer specifies a desired sequence structure showing the sequence relationships between conflicting traffic streams. Normally, each stream appears once each cycle. An exception is free realization, in which a tram, bus lane, or other priority stream with infrequent requests and a small need for green time can be inserted anywhere in the cycle (though not until conflicting streams have become red). Free realization streams do not appear in the sequence structure. It is also possible to have selected streams appear more than once per cycle at specified locations in the control structure, although this option is not discussed further for lack of space. An intersection can have a large number of possible sequence structures; some are better than others in terms of critical cycle length, lost time, efficient use of extra green time, providing good coordination where desired, and so forth. Several authors (e.g., 11, 12, 13) have explored the subject of finding the best sequence structure for an intersection under fixed time control with a time-independent arrival process. Many of the same principles apply for actuated control as well, although Bell has rightly observed that the optimal sequence when traffic patterns are assumed known may not be optimal when they are random (14). This paper takes the sequence structure as given. Cyclic sequencing be contrasted with one a-cyclic method used in the Netherlands, in which streams are in turn based on earliest request. While there are some advantages to first in – first out logic when demand is light, its drawback is that when demand becomes heavy, the operation inevitably becomes cyclic, and the signal program may then be locked into an inefficient sequence for a long period of time. A desired sequence structure can be visualized by means of a structure diagram in which streams are arranged vertically in the desired sequence (downward with time). It repeats in principle without end, although for practical purposes, a little less than two full cycles is usually adequate to illustrate all of the stream relationships. A rectangle represents a stream’s active period, and arcs show sequence relationships. Conflict arcs connect the end of one stream’s activation with the next activation of a conflicting stream. Coordination arcs, represented by double lines, connect the activation start of simultaneous start streams. Staggered start constraints are also represented by coordination arcs, which connect the later stream’s activation start to some point during the former stream’s activation. To reduce clutter, the streams that lie in a column form a conflict group, meaning that they are all in conflict with each other, so that conflict arcs between streams in the same column are implicit. Example Intersection To illustrate these points, an intersection is sketched in Figure 1. Streams 35 and 36 are pedestrian streams; stream 46 is a tram/bus stream in its own right of way; the remainder are general vehicular streams (numbered according to the standard codes common used in the Netherlands) Figure 1. Example intersection Sequence structure Flow diagram A sequence structure and corresponding flow diagram (stage flowchart) that will be used in later examples is also given. Some optional tactics for this layout, not incorporated, include: conditional free realization for tram stream 46; truncating other streams’ green upon a request from stream 46; simultaneous start for opposing streams 5 and 11; having pedestrian stream 36 precede stream 11 by, say, 2 s; and providing green waves through pedestrian streams 35 and 36, with the direction of the wave dependent on which pedestrian pushbutton was actuated. Sequence Control and Sequence State The problem of sequence control is to ensure that the signal program advances in a manner consistent with the specified sequence structure and priority tactics. When there is a fixed sequence of stages with no priority interruptions (as in fixed time 02 03