discussIT.org
to Collaboratively Create Authoritative Works by Optimizing the balance between Exclusivity and Openness

 

2006 TRB Annual Meeting paper with additions and sections that were eliminated because of length restrictions.

 

Roadtrains


Abstract

All-electric vehicles might be physically coupled to form “roadtrains” led by human “pilots”.  These could triple lane capacity on average (quintuple with staging in advance of lane reductions), double fuel economy and facilitate electrification of the roads.  By taking advantage of with robotic driving at low speeds, roadtrains have potential to greatly increase ridership on public vehicles.  These substantial benefits motivate the hypothesis that roadtrains are a practical solution to future transportation challenges.

Discussion about how roadtrains might form, operate and disperse also considers backward compatibility so that roadtrains can operate in mixed traffic and with varying degrees of automation, route communication and planning.  Factors that positively and negatively impact safety are considered and suggest ways in which roadtrains could significantly reduce risk relative to solo cars.  Institution of roadtrain priority lanes to reduce travel time is suggested as a primary driver for adoption.

This paper does not conclusively confirm the hypothesis but should serve to open a discussion that could become particularly active in the context of the many supporting technologies now becoming available.

Introduction

Background

Transit Ridership Must Increase

Mass transit is often proposed as a solution to congestion and the threat of diminishing oil supplies but transit (including rail) accounted for only 1.01% of surface person-miles in 2003 in the United States.  The term ‘mass transit’ is self-explanatory with respect to this failure.  Cars are the transit of the masses.  ‘Mass transit’ is a correct term only in the sense of being able to transport a massive number of people in one vehicle, be it a bus or railcar.  Yet these large vehicles perform poorly on three key attributes: the ability to provide on-demand, non-stop and door-to-door service. 

    To become desirable, transit must compete on the elapsed time between demand and delivery.  This can only happen if the vehicles are reduced in size by an order of magnitude.  Without automation, this radical re-conception of transit is impossible because driver costs – some 60% of operational expenses – would also increase by an order of magnitude.  The CyberCar (2) project has demonstrated that fully automated vehicles can provide public service along restricted corridors at low speeds.  One of the lead researchers, Michel Parent of INRIA, stated in a personal conversation that enabling the vehicles to physically connect is an approach worth investigating.  This could allow small, quasi-personal and potentially driverless vehicles to assemble into trains led by a human at higher speeds on major arteries.  Dr. Parent also alluded to substantial safety advantages relative to close following. 

Problems of Automated Platooning Vehicles

Automatic close following prototypes have been developed for trucks in the CHAUFFEUR project (4) and cars in the Automated Highway System (AHS) (5).  Both operate with substantial gaps between the vehicles and will not be deployed unless barriers separate them from regular traffic.  This poses a substantial “chicken and egg” problem (6): setting lanes aside only for automated use before people get automated vehicles would result in empty lanes but people won’t buy the vehicles unless the lanes are available.

There has been considerable discussion about building towards an AHS solution incrementally.  Koopman and Bayrouth (7) define four baseline capabilities that can be implemented independently: free agent versus platooning, single versus multiple lanes, obstacle exclusion versus detection and avoidance, and no system vigilance (trust) or complete system vigilance (vehicle’s complete distrust of the system and other vehicles).  Tsao (8) suggests a strategy of introducing AHS capabilities into mass transit vehicles for use on existing high-occupancy-vehicle (HOV) lanes, subject to the supervision of a safety driver.  Both approaches are criticized by Shladover (9):  Koopman et al and Ward (10) for ignoring the implementation difficulties of a fully automated system; Tsao’s automated shuttle-bus with driver back-up, because of the partial attention problem studied by Hogan (11).  

The main conclusion in the summary of the final report on the Automated Highway System states, "The demonstrated technologies (platoons) ... could not resolve the complex problem of allocating these increased traffic volumes safely and efficiently into the traffic streams of already congested local streets"(5).  Segregated and barrier protected AHS-type lanes cannot generally be used on city streets. 

In conversation with the author, Shladover posited that traffic flow on city thoroughfares is too hazardous for platoons because of the likes of cyclists and opening car doors.  Long trucks and buses routinely use city streets so the problem is not vehicle length as long as the vehicle is continuously connected.    The problem is the inability of automated driving systems to deal with highly complex environments.  This suggests that fully engaged human drivers are needed in the urban core and at higher speeds. 

A New Kind of Roadtrain

In Australia, roadtrains consist of trucks with two or three trailers that are hitched manually for long trips across the Outback.  The term ‘roadtrain’ is thus used in this paper to distinguish physically connected vehicles from those that are not (e.g. AHS and CyberCar platoons).

This paper suggests a second definition of roadtrain wherein the following vehicles still have physical connections and a fully engaged and specially licensed human driver but differ in other characteristics: the vehicles are usually used for passenger transport, have their own powertrains, can operate individually and tend to be highly automated and sensor equipped.  Like their Australian cousins, the trains may be led by massive tractor vehicles with freight trailers.

The term “traincar” shall denote any following vehicle in a roadtrain.  They may have a variety of purposes and levels of automation. 

Advantages of Mechanically Hitched Roadtrains

Energy Efficiency and Comfort

Traincars may be designed to mate aerodynamically.  This will allow relatively large and comfortable vehicles to achieve excellent fuel economy.  The potential is suggested in a study by Zabat et al (16).  In this wind-tunnel study, scale models of four Chevrolet Lumina vans operating at zero following distance were used to show 57% and 59% reductions in aerodynamic drag for vehicles in the 2nd and 3rd positions.  The first and fourth vans showed drag reductions of 40%.  These numbers are too optimistic because the scale models lacked rotating wheels; too pessimistic because the shape of a Lumina van is not optimized as a traincar.  At .30, the coefficient of drag of a Lumina van is quite similar to conventional sedans (17) so the gains from platooning are likely to be similar.  For the subsequent estimation, the average aerodynamic drag in a 5 car platoon is given as 49% that for a single car ((60+43+41+41+60)/5).

Fuel economy will also be boosted because the trains would be naturally hybrid-electric.  Vehicles used exclusively within and between metro areas can be all-electric yet enjoy unlimited range.  Their batteries need only provide independent power for a few minutes at freeway speeds when merging or changing locomotives, for low speed neighborhood service and climbing modest hills.  (Lane throughput will be affected by the acceleration characteristics of the vehicles so peak output may be augmented by a source of burst energy such as super-capacitors.)  Thus, most of the roadtrain fleet would avoid the weight, size and cost otherwise imposed by Internal Combustion Engine (ICE) engines and their transmissions. Hence, the weight of the all-electric car is estimated to be 60% that of a comparably sized conventional car.  This reduces the rolling resistance which represents 25% of the total at 120kph (73MPH).

While the followers can use regenerative braking to save energy, this advantage is not so important with future traffic control systems that advise about over-acceleration.  The ability to operate the locomotive engines nearer to their most efficient operating point will probably produce more important gains.  Typical cars now cruise inefficiently at around one-tenth of their maximum power with very large reserves of torque at typical engine speed.  One may reasonably expect highway fuel mileage gains larger than 2X while maintaining interior volume (1/((.75*.49 + .25*.6)*.85)=2.27) where .49 is aero drag reduction, .6 weight and .85 more efficient use of the engine.  This suggests that the EPA highway rating of a Lumina sized vehicle could go from 26MPG to 59MPG.  Aerodynamics has little effect on city mileage ratings so the improvement factor is likely to be smaller (especially if compared to hybrid vehicles or future cars that pair idle engine cut-off with acceleration cut-out based on advance communication of signal changes) but the weight loss and more efficient engine operating point still suggest that an improvement factor around 2X.

Electrification

From the perspective of a national strategy to reduce dependence on oil, the most important benefit comes from electrifying the arterials.  Overhead wires used by buses and light rail today must allow the tallest trucks (13’ 6”) to pass underneath.  Electric pick-ups able to reach elevated wires (and possibly deploy automatically) would be costly if mounted on every vehicle.  Instead, roadtrain locomotives plying these routes could be equipped with pick-ups that power an entire train. 

Congestion Relief

A primary driver for mass transit adoption is congestion relief.  Right-sized 11’ long traincars (two-thirds the length of today’s cars due to space savings from eliminating the engine) would be able to increase vehicle lane throughput owing to reductions of both car length and average following distance.  At 100% adoption, freeway lane capacity now rated at 2,200 vehicles/lane/hour may be expected to improve by an average factor greater than three.   25’ locomotives pulling an average of five 11’ traincars at 60MPH with two seconds separation between trains yields 7,425 vehicles per hour. (5280*60)/(25+5*11+5280*2/60)*6) 

If the length of the train is increased to 3 locomotives and 15 traincars with three second separation – this might coordinated in advance of lane closures for construction zones or chronically overloaded sections (such as during the peak of rush hour or a mass evacuation) – lane capacity could be increased to 11,314 vehicles/hour.  (5280*60)/(25*3+11*15+5280*3/60)*18)

In the city, traffic signals limit throughput.  Roadtrains will greatly increase capacity by eliminating not only the gaps between vehicles but most of the cumulative delay before accelerating.  Moreover, roadtrain corridors would certainly employ adaptive and cooperative signaling.  Since roadtrain vehicles will reduce the frequency of individual vehicles, it will be easier to arrange continuously green lights for the trains.

Convenience

Whether the traincars are fully or partially automated, most of the transit time will be in follower mode which allows drivers to be relived of their responsibilities. 

On door-to-door (D2D) trips, roadtrains will provide time-to-destination similar to conventional cars operating on the same streets.  On account of their high efficiency and lane throughput, roadtrains may be granted privileged access to lanes.  These can markedly reduce trip duration - dramatically so if roadtrains are also given signal priority. 

Operation of Roadtrains in Mixed Traffic

Need for Uncoupling and Coupling while Moving

Roadtrains should be able to uncouple while moving so that following vehicles can exit the roadtrain without forcing the entire train to stop.

Although stopped vehicles could couple into roadtrains near freeway entrance ramps or at designated areas similar to bus stops, this approach would require land and development.  It would also require roadtrains to make stops in order to add followers.   (While coupling could also occur when cars are stopped in front of red lights, such opportunities should be rare under the cooperative signaling to be expected on the roadtrain routes.)

The following discussion therefore addresses the assembly and dispersal of moving roadtrains; the case of a coupling with a stopped roadtrain is a trivial extension.

Need for Docking ACC (D-ACC)

During docking, a car’s following distance decreases from a safe headway for a human driver to zero.  During this period, drivers must cede longitudinal control to coordinated automatic systems that finely control the motor and can apply maximum braking if necessary.

The mathematics of platoon joining and dividing is well understood (12) and the techniques for Cooperative Adaptive Cruise Control (C-ACC) much discussed.  Roadtrain compatible cars would require an upgrade to permit actual docking.  This may come to be known as Docking ACC (D-ACC).  As with ACC, D-ACC would maintain speed until the radar detects another car ahead in the same lane.  If the mating criteria below are met, D-ACC would reduce the following distance until coupling is achieved.  During the last few meters of approach, an additional vision based sensor may be needed to measure the gap size with sufficient resolution.

Lateral Control during Docking

Human drivers should be able to reliably steer the docking mechanism into position.  If not, final steering can be automated with a low cost camera and suitable markings below the leader’s bumper.  Haptic feedback to the steering wheel might let the driver know when the system has taken control in the final moments before contact (13). 

The lead vehicle may employ a Lane Keeping System (LKS) (15) when followers are docking.  Such an LKS may be de-activated after docking so that the lead driver remains attentive. 

We should also consider the case where the lead vehicle makes a sudden evasive turn to avoid an obstacle.  To address this case, the following vehicle may employ haptic steering that becomes progressively more forceful as the following distance is reduced.  This would help the following driver to track the lead vehicle rather than be exposed to a suddenly revealed hazard.

Coupling Rules and Etiquette

This section details how human drivers couple with a suitable roadtrain.  The same protocols may be encoded in software.

Joining at the Rear

Additional followers will tend to decrease the risk for the traincars ahead by increasing visibility of the train.  Other than ensuring that the locomotive’s actual towing capacity is not exceeded and that the vehicles ahead are of similar or heavier weight class (see Sequencing below), there are no restrictions or coordination procedures.  The joiner simply enters the lane directly behind the roadtrain, engages D-ACC and couples.

Merging

With two or more unidirectional lanes, proper merge procedure consists of:

1.      Pulling alongside a roadtrain.

2.      Signaling the intention to merge with the turn signal.

3.      Looking for someone in the adjacent roadtrain to create an opening and adjusting speed to move parallel. 

4.      Engaging D-ACC (if this did not automatically happen in step 2). 

5.      Waiting for positive acknowledgement from the D-ACC that the vehicles ahead and behind have electronically linked longitudinal controls.  The merge can now occur safely in a very small gap. 

6.      Changing lanes to merge and couple.

Preparation for Uncoupling

Drivers who are exiting should give following drivers sufficient time to orient as to location and determine whether to follow or uncouple.  Their follower(s) may also need time to disengage D-ACC, establish a safe following distance and resume control.   

Roadway signage and markings may need to be displayed or spoken to the following drivers because they have little or no forward vision.  On freeways, several miles of advance notice may be needed to permit roadtrains to smoothly regroup. 

Uncoupling

Before the traincars uncouple, the driver prepares to take control and disengages D-ACC which uncouples from the vehicle ahead and starts opening a gap using cooperative ACC.  Because visibility is limited just after separation, it will be very important to closely follow the path of the car ahead until the sightlines provide adequate headway.  When separation grows to be compatible with the driver’s reaction time, cruise control can be manually disengaged.

Uncoupling can shift leadership to an automated system or someone who may not be certified as a roadtrain pilot.  These will be periods of elevated risk and should be kept as few and short as possible. 

For Other Vehicles

Vehicles that use much more road space and fuel may be prohibited from obstructing roadtrains or completely excluded in certain lanes. 

When only a single lane is available and entering traincars have no acceleration lane, other vehicles may be required to leave enough space behind a roadtrain for entering traincars to enter the lane, accelerate and join the roadtrain. 

For Leaders

When departing, pilots-in-command must give adequate notice in the same manner as an uncoupling follower.  When multiple lanes are available, exiting roadtrains move to the exit side and reduce speed far in advance.  

Leaders should use speeds and lanes that facilitate the switching of traincars.  This means that local roadtrains operate in the center or outer lanes at speeds which cause long distance roadtrains to pass.  These passing events will provide opportunities for traincars to switch to roadtrains going to more distant exits.  (With Digital Short Range Communications (DSRC) between navigation equipped vehicles, the speeds and switching operations can be choreographed by on-board computers.)

When they are not needed to lead, locomotive pilots should join existing roadtrains in order to free up road space and maximize safety and energy efficiency.

Leaders may also be required to “give clearance” to automated vehicles attempting to join or leave their roadtrain.  This action consists of observing the approaches that an automated vehicle will be using during their acceleration and merge or departure and deceleration.  If the approaches are not clear, it may be the responsibility of the leader to either slow to a speed which is safe for the affected vehicles or, if this is not possible, to abort the operation.

For Waiting Traincars

All-electric traincars will often require a locomotive for their journey.  Since no time is saved by entering a thoroughfare before the locomotive has passed, they should wait and will need to take care not obstruct other traffic.  (This may require bus stop like waiting zones on the thoroughfares and their feeder streets.  Inter-vehicle radio communication can reduce the need for waiting zones by adjusting the speed of joining traincars so that they arrive just in time to merge into a roadtrain on the thoroughfare.  Radio coordination could also allow joining vehicles to remain parked at some distance until a roadtrain approaches.)

Sequencing of Traincars

Heavier Traincars Generally Frontmost

Heavier cars generally go in front because:

1.      Heavy traincars are best able to withstand a primary collision.

2.      Heavy vehicles are more likely to have reserves of power sufficient to act also as locomotives.

3.      Bigger vehicles tend to seat the driver higher which affords a better view and anticipation of conditions ahead.

4.      Lightest-last configuration reduces crushing forces on cars in the middle of the roadtrain during a severe crash.

5.      Followers will be most stable in case of a sideswipe if they can remain connected and stay more or less on their leader’s path.

Exceptions from the Heaviest First Rule

The heaviest foremost rule need not be absolute.  A tapered nose (and tail) provides the best aerodynamics. Thus risk affine drivers of such sporty cars may be allowed to drive in the lead even if heavier vehicles follow.  This configuration can improve overall safety by affording dual piloting.  In this case, the first vehicle would have a co-pilot while the driver of a taller following vehicle would look over the smaller vehicle and be able to override the co-pilot but not steer his vehicle.  In the event that steering is overridden by the pilot, instantaneous disconnection from the smaller car ahead and separation by braking would ensue.

If no trucks with qualified drivers are available to serve as locomotives and lead vehicles, the reserve power of standard sized cars may be used.  The followers’ combined weight can substantially exceed the rated towing capacity because they can draw on batteries to accelerate independently and climb modest hills.  

Courtship of Roadtrains

On the multi-lane thoroughfares where roadtrains will be primarily deployed, the inner lanes will carry through traffic and the outer lanes slower moving local traffic (except when exits are on the left).  Traincars entering the thoroughfare will usually attach to the next roadtrain.  If it is not going as far as desired, the driver can take the next opportunity to switch to a faster moving train on an inner lane.

            Communication of routes can reduce the number of switching events.  While this is highly desirable and also rather likely to be a standard capability by the time that traincars can be deployed, it is worth noting that that roadtrains may be formed by incremental changes to familiar driving behavior. 

Roadtrains generally continue in the same lane until exiting so once a traincar has found a pilot going at least as far, there will be much less potential for lanes changes made by individual drivers (weaving). 

Mating with no Prior Information

When no destination information is available, cars will simply join at the end of other cars having the same weight class.  As the roadtrain progresses, this will tend to sort the more distantly destined vehicles so that they end up near the front where their passenger-drivers are not inconvenienced by the comings and goings of local traffic.

Turn Signal Only

In the most basic implementation where there is no route communication, courtship is limited to matching vehicle class according to mass and may be determined by visual inspection.  Turn signals are a minimal but backward compatible and sufficient way to indicate a desire to merge into a roadtrain or exit from one. 

Exchange of Intended Routes

Intermediate traincars can exit only by splitting the roadtrain.  This entails some additional risk and may also be inconvenient to the following driver.  Thus it is desirable to sequence the traincars according to destination in order to minimize the number of uncoupling operations and lane changes.  In this way, the next traincar(s) to exit should also be the last traincar(s).  Various means of communicating routes may be employed.

Intercom  An intercom designed to only permit communication between prospective and approaching traincars headed in the same direction could be used to communicate destinations.  Once in a roadtrain, drivers would not have to monitor communications if the intercom could be interrogated to replay a recording of the planned exit.  The intercom may also be used to in lieu of signage that cannot be seen due to loss of forward vision. 

Navigation System  As navigation systems become more commonplace and able to predict routes based on history, computers will handle more and more of the route matching communications chatter.  This will reduce demands upon drivers such as decisions about when to enter the thoroughfare, which lane to use and when to join or depart from a roadtrain.  It is quite likely that this will be the only type of system most users would encounter.

Relieving Pilots

Pilots should be periodically relieved in order to maximize safety.  “Ready for relief” status should be automatically raised by measures of drowsiness such as lane drifting and eyelid closing.

The pilot will usually request a change of status manually or be reminded by to do so by a timer that counts down pilot-in-command time.  The pilot continues until a suitable replacement acknowledges the call by merging in front.  The driver being relieved may either engage D-ACC or uncouple. 

Other Factors in Matchmaking

Drivers and their automated agents may negotiate other factors such as price, aerodynamic compatibility and safety related considerations such as driver/vehicle rating and dispersing alert drivers throughout a roadtrain.

Characteristics of the Physical Coupling

In order to allow for high frequency lateral motion, the female half of the coupling will need a guiding shroud.  The following traincar may insert a necked rod that is threaded into an assembly with a spring, damper and a mechanical linkage to the steering.

Coupling Strength

The width of the neck on the connecting rod can determine the strength of the coupling.  It must be strong enough to keep the roadtrain connected and stable in case of road surface irregularities (e.g. potholes), tire puncture and even wheel loss.  Physical connection will tend to stabilize traincars experiencing collisions with other vehicles traveling in the same direction because the traincars ahead in the roadtrain will pull the affected traincar back into the center of the lane. 

Automatic Uncoupling and Defensive Maneuvers

The connecting rod should disconnect when the vehicle ahead receives a heavy side-impact. 

Following traincars should be programmed to minimize or avoid involvement by applying maximum brakes.  This would divide the platoon at a time when drivers in the following traincars are likely to be inattentive.  Best outcomes will require vehicles to employ Lane Keeping Systems and automatically engaged defensive maneuvers.  Optimization of these maneuvers will prove to be a rich area for R&D.

Steering Linkage, Turning Radius & Path

In order to minimize risk of failure, a redundant mechanical steering linkage is likely to be desirable. 

The usual mode of roadtrains is to continue straight in the same lane.  However, it will be desirable in some heavily trafficked areas to be able to make sharp turns (e.g. from one right hand lane to another) without having to uncouple.  During such small radius turns, traincars will need to be able to yaw substantially relative to the car ahead in order to follow precisely.  This may excessively complicate the coupling design and either necessitates a sharp turn prohibition or else a routine that lets cars briefly uncouple and turn independently.

Mechanical linkage may be eliminated if a tri-wheel design is used where the third wheel retracts.  Yaw might then be controlled through independent motors for the remaining two wheels.  This design would also have the advantage of eliminating one-half of the rotating wheel drag. 

Measurement of Tension and Power

Each coupling could have tension sensors used to calculate the power supplied by the locomotives.  These may be re-calibrated in every roadtrain by reference to readings from the other couplings.  Consistent under-reporting of the tension would then occur only as a result of illegal reprogramming.  Any such attempts to underestimate power consumption can be easily caught by reporting instances of significant inconsistency in tension readings. 

Factors that Raise collision risk

Impacts Propagate Back to Followers

Cars driving with normal headway are not nearly so likely to become involved in a collision when the car directly in front collides.  This increased vulnerability must be addressed with prevention and impact minimization strategies.

Unplanned Lane Changes More Likely to Cause Sideswipes

Lane changing to avoid direct collision is more likely to cause side-on contact with adjacent vehicles because of the length of road-trains.  These evasive actions will be justified when they cause less harm than the alternative frontal collisions.  (However, the damage will be sustained by innocents rather more than the party at fault.  Lead drivers may elect to have their vehicles equipped with “black box” recorders that include forward facing video so that these accidents can be reviewed and fault assessments properly made.)

Factors that Lower collision risk

Reduced Roadway Departure and Rear-end Collision Risk

Roadway departures and rear-end collisions can be reduced by mandating that:

1.      Driving in the lead is a special privilege (and a possible source of earnings).  It may require a commercial type license with especially stringent testing criteria.

2.      The lead vehicle should have warning systems such as lane departure, forward collision and driver alertness.

3.      Poor driving is immediately reported.  Passengers in the following traincars will usually notice erratic driving and can opt to uncouple.  Unscheduled exodus can be automatically reported and may result in loss of leadership privilege.

Decreased Danger for Other Road Users

Roadtrains will tend to decrease collision risks and consequences for other users of the roads.  A zone of danger exists in front of a moving vehicle with an area that expands exponentially with speed.  Roadtrains reduce the number of these zones by aggregating many vehicles into one.  Pedestrians, cyclists, people opening car doors and animals are virtually guaranteed to be less injured if they collide with the side of a passing roadtrain than if they are struck by the front bumper of one of its constituent cars operated individually. 

Other users of the road are less likely to enter the danger zone or hit the sides of the vehicle because roadtrains will be much bigger and more visible than individual cars.  (The smooth streamlined sides of the traincars should have no protrusions likely to cause serious injury.  External mirrors may be replaced with sensors or video cameras, or made out of lightweight materials and designed to fold flush with the surface if they strike something.)

Minimizing Risks of RoadtrainS

Active Intersection Control Required

Several technologies to prevent intersection accidents are under development.  These divide into two types: 1) roadside detection of vehicles at risk of unlawfully entering the intersection against a red light or without stopping, 2) radio based communications between vehicles providing location and speed that allow vehicles to independently determine if there is a significant accident risk.  Both are an added layer of security on top of existing traffic signals and signs so provide benefits even if only partially deployed and effective.

One-way or Barrier Divided Streets

Head-on collisions with oncoming vehicles are relatively rare on one-way streets.  Unless most other vehicles and drivers can be brought under strict standards to ensure lane keeping, roadtrains might be restricted to one-way streets.  If used on two-lane streets, travel in the oncoming direction may be restricted to vehicles with employing special safeguards against left-turn violations.

Support for Safe Lane Changes

As with long trucks, lane changing by roadtrains will be limited by the higher probability of adjacent vehicles.  Lane-changing cannot be strictly avoided so it will be very important to prevent unsafe lane changes.  This may be achieved by a combination of adjacent and approaching vehicle detection, cooperative signaling, early advisories of upcoming lane changes, and laws that reduce the probability of unsafe conditions.

Adjacent vehicle detectors are already available and are used to alert the driver when the vehicle is on a collision path with an adjacent car or nearby object.  These may be augmented by rearward facing detectors to detect approaching vehicles.

Radio communication is expected to be used to alert drivers of upcoming lane closures and blockages due to construction, accidents and congestion.  These can be propagated such that roadtrains are able to enter the correct lane well in advance of any lane closures.

Emergency lane changes present the greatest risk for collision with adjacent vehicles.  One mitigating factor is that the roadtrain need not translate sideways when changing lanes but can change lanes in a snakelike fashion with each car following the one ahead.  If they are not near the front, this gives adjacent drivers additional time to maneuver.

Enforcement of “No passing on the right” and “Slower traffic to the right” laws can also help to prevent riskier conditions.  It would also be helpful if roadtrains normally have right-of-way over other vehicles.

Detection of Hazards

Road debris large enough to damage and divert cars must be removed from the roadway as quickly as possible.  This is important for conventional cars but especially important for roadtrains because their ability to swerve is more constrained by the likelihood of vehicles in adjacent lanes.  Debris also tends to accumulate between lanes where it can be especially hazardous to traincars that are switching in heavy traffic when they might not establish sufficient forward visibility to be able to avoid striking the debris. 

Radio communications otherwise used to communicate hazardous conditions such as road icing and accidents may also be used to alert approaching vehicles and even stop them so that the condition can be remedied. 

Designing for Collision Safety

Traincars behind the point of collision will benefit from an extended crumple zone consisting of in-between traincars.  The further a traincar is back from the point of impact, the less the damage.  However, the crumpled cars should not create unnecessary casualties caused by crushing the occupants.  Just as cars today are designed to deflect the engine downwards when hit head-on, roadtrain cars must be designed to preserve the integrity of the cabin by deflection rather than allowing the roadtrain to collapse in an accordion-like manner. 

Prevention of Jacknifing

During normal operation, the couplings will be kept in tension because the follower cars are being pulled.  When the leader brakes, the following cars should brake so as to maintain tension.  This may be achieved by sensing displacement of the attaching spring and recalibrating the brakes every few milliseconds.  All vehicles and tires should have similar stopping specifications.

Compression of the couplings will still be inevitable under hard braking if trailing cars in the roadtrain are on patches of water, oil or ice.  This relatively infrequent occurrence can also cause compression contact and controllability problems known in the trucking industry as “jackknifing”.

Such situations call for automatic differential braking, steer-by-wire and possibly even automatically reducing braking in the lead vehicle if no collision risk is sensed.  Work is needed to make these systems highly effective and also to broadcast such situations to approaching vehicles so that they can adjust their speed accordingly.

Active Protection in Crashes

Pre-crash sensing may also determine that contact is inevitable and guide the traincars using data available to predict body dynamics.  This data may come from a variety of sources that include video from the lead vehicle and trajectory predictions provided by the vehicles ahead.  In an emergency, the communication subsystems should be able to exchange this information in a timely manner.

Optimizing Roadtrain Length for Safety

The major safety advantages of roadtrains (less frontal area, replacement of unfit drivers, use of advanced driving aids and monitoring in lead cars, having vehicles ahead to cushion impact) reduce risk as a function of the number of coupled traincars.  The major safety disadvantages (involvement of all following cars in serious collisions and increased sideswipes when steering to avoid frontal collision) increases risk with the number of cars in the roadtrain. 

These produce oppositely sloped curves for the risk of operating a roadtrain as a function of its size.  The minimum risk may be found by summing the curves.  However, it may be found that the negative slope of the risk reduction curve always exceeds the positive slope of increasing risk with the number of traincars.  In this case, even though the risk that vehicles ahead are involved in an accident increases, the collateral damage from such an impact drops off sharply in following vehicles.  This could make the rear cars of long roadtrains most safe.  (It is conceivable that the levels could be comparable with rail safety and make it possible to produce sleeper traincars.)

The risk curves will change according to the driving environment.  Along remote sections of interstate highway with light traffic and large medians affording negligible risk of head-on collisions, a roadtrain with dozens of traincars may be considered safe.  (Locomotive power limits length but an electrified highway could allow traincars to largely self-propel.)  Roadtrains will tend to be shorter in dense city traffic with many intersections and cars in adjacent lanes.  They will be shortest or non-existent if the roads are two-way without median barriers.

More detailed risk modeling is needed to inform safety engineers.  Once roadtrains enter service and begin having crashes, detailed data should be collected and analyzed to minimize risks.

Of course, it will be up to each driver or vehicle to use this data when deciding whether to join a roadtrain.  A poor driver or a drunken one will be much safer in a traincar but still choose not to use one. 

Psychology

There is always resistance to change and drivers will have specific objections to using roadtrains.  The sensation of inadequate headway during docking will be unnatural and frightening to some drivers.  This may be overcome with D-ACC experience that may be acquired on transit systems and rented vehicles.

Trust will need to be established and this must be achieved first and foremost by a record of safe operations.  Major improvement relative to solo cars will be needed to overcome the natural reluctance to cede control to an unknown pilot. 

Driving effort will be greatly reduced because most users will be followers with greatly reduced responsibilities.  The lowest level of automation (D-ACC with communication via turn signals) requires only an incremental adjustment to current driving behaviors.  Optimization of roadtrain selection using an intercom will add complexity but is likely to be optional and overleaped by on-board navigation systems that reduce demands on the drivers.

The perceived loss of freedom due to giving up control is a recurring concern.  Where driving is viewed as an opportunity for expression of a competitive personality, the conversion can be challenging.  Driving depends much more on cooperative than competitive instincts and the key to socializing drivers is rewarding the former rather than the latter. 

Motivating Adoption with Priority Lanes

The traveling public is very strongly motivated by time savings.  Dan Kirshner, President of Ride Now, Inc stated that the 15-20 minutes saved (by being able to use the HOV lane on I-80 east of San Francisco) is by far the dominant reason that people choose to change their commuting behavior.  This suggests that priority lanes on both congested highways and city thoroughfares could strongly encourage people to use roadtrains. 

The time savings of such priority lanes could be greatly amplified with signals that also give priority to roadtrains.  (Signals that are at least aware of longer roadtrains will be needed to ensure that the entire train clears the intersection before cross-traffic receives a green.)

Such roadtrain priority lanes will also need to be open to roadtrain compatible locomotives even if there are no traincars to pick up.  This will be very important to initiating the adoption process.  While the majority of vehicles used in cities should be all-electric, these will depend upon locomotives for trips of more than a few miles.  Thus locomotive capable vehicles must have already captured a substantial market share in order for the all-electric models to have broad appeal.

Once roadtrains have been well established, there should be a small surfeit of locomotives relative to follower vehicles.  The correct ratio may develop out of financial incentives to participate in relatively efficient roadtrains and competition to provide piloting and locomotive services.  Payments can be made electronically by means of encrypted electronic cash that does not require real-time communication with a bank.

ConclusionS

Operation of roadtrains on streets with mixed traffic presents some additional operational complexity.  However, automation such as Docking ACC will make the changes incremental and manageable from the driver’s perspective.  On-board navigation and communication systems will be able to handle almost all of the decision making about synchronizing with roadtrains.

Using planned safety enhancements such as intersection collision avoidance using Digital Short Range Communications (DSRC) and lateral and rearward vehicle sensing, roadtrains should be able to demonstrate significantly improved overall safety.  This improvement will come principally from reducing the effective number of vehicles on the road and having many fewer unsafe drivers operating at higher speeds.

100% adoption of roadtrains would likely increase roadway throughput 3X.  Their streamlining and hybridization could more than double the fuel economy of independently operated, similarly sized cars.  These substantial public benefits should justify roadtrain lanes with signal priority.  These could provide very substantive time savings that strongly stimulate private purchase of roadtrain compatible vehicles.  Physical coupling to combustion powered or electric locomotives should allow follower vehicles to have very low complexity, operational and purchase costs. 

Electric locomotives could also lead to reduced dependency on oil by instead using proven non-oil based fuels and even some renewable sources of electricity.

Roadtrains facilitate a separation of higher speed driver mandatory operation from low speed fully automated driving.  Transit systems could purchase fleets of right-sized buscars able to autonomously transition between roadtrains and elongated bus stops.  A prevalence of locomotives would allow transit users to reasonably expect nearly on-demand departures, few stops and expanded routes.  Some traincars could autonomously rove neighborhoods at low speeds to provide taxi-like service.  Such golf cart-like electric cars would also allow short term car rental agencies (e.g. ZipCar and FlexCar) to relocate vehicles automatically which would help them to offer much more affordable one-way leases.  The increase of transit service to private car levels and the decrease of rental car pricing towards transit levels should lead to beneficial reductions in private car ownership, especially in the cities.

Roadtrains would also greatly facilitate realization of fully automated parcel and freight services that lower delivery costs and accelerate the Internet economy.

The wide ranging set of advantages suggests that substantial research and development effort is merited.  Safely coupling moving vehicles and designing roadtrains to be redundant, failsafe and able to sharply reduce damage in traincars behind the point of impact are among the foremost technical challenges. 

Acknowledgments

I would like to thank Steve Shladover and Michel Parent for sharing some of their wide and deep experience in the field.  I would also like to thank Rodney Lay and the expert panelists at the Future of Roadtrains special session at the ITSA World Congress 2006.  Wei-bin Zhang’s observations were especially relevant.

I am ever thankful to Jerry Schneider for his selfless and decade long pro bono efforts to maintain a web site on innovative transportation technologies.  Among the numerous inventors cited there, I thank Charl du Toit (3) for exploring the use of physical couplings to extend the range of electric cars and to reduce congestion.

The support of Christine Economides, Jim Longbottom and the Richard Lounsbery Foundation has helped to make this work possible.

rEFERENCES

1.      Polzin, S. E., Observations About Public Transportation Based on NHTS Analysis. National Household Travel Survey Conference: Understanding Our Nations TravelWashington, DC  November 1-2, 2004.   

2.      Cybernetic Technologies for the Car in the City Homepage.  http://www.cybercars.org/.  Accessed July 31, 2005.

3.      FlexiTrain.  http://www.camdek.com.  Accessed August 1, 2005.

4.      CORDIS.  CHAUFFEUR TR 1009: Summary of Chauffeur project, Schulze, M., project coordinator http://www.cordis.lu/telematics/tap_transport/research/projects/chauffeur.html .  Accessed July 31, 2005

5.      Review of the National Automated Highway System Research Program. In Transportation Research Board, TRB Special Report 253, TRB, National Research Council, Washington, D.C., 1998.

6.      Tsao, H. and S. Jacob. Axiomatic Approach to Developing Partial Automation Concepts for Deployment of Automated Highway Systems and Partial Invocation of Vision-Based Lane -Keeping and Adaptive Cruise Control.  In Transportation Research Record: Journal of the Transportation Research Board, No. 1651, TRB, National Research Council, Washington, D.C., 1998, pp. 74-79. 

7.      Koopman, P. and M. Bayrouth.  Orthogonal Capability Building Blocks for Flexible AHS Deployment.  ITS Journal, 1997   

8.      Tsao, J.  Constraints on Initial AHS Deployment and the Concept Definition of a Shuttle Service for AHS Debut.  IVHS Journal, Vol. 2, No. 2, 1995, pp. 159-173. 

9.      Shladover, S. E.  Progressive Deployment Steps Leading Toward an Automated Highway System.  In Transportation Research Record, No. 1727, TRB, National Research Council, Washington, D.C., 2000, pp. 154-161 

10.  Ward, J. Step by Step to and Automated Highway System-And Beyond:  In Automated Highway Systems.  P. Ioannou, ed., Plenum Press, New York, 1997, pp. 73-91. 

11.  Hogan, R. M. Impact of Physical Disengagement on Driver Alertness: Implications for Precursors of a Fully Automated Highway System. Proc., IEEE Intelligent Transportation Systems Conference, ITSC-97, Boston, 1997.

12.  Horowitz, R. and P.Varaiya. Control Design of an Automated Highway System.  IEEE, Proceedings of the IEEE, Vol. 88, NO. 7, 7, July 2000

13.  Lateral Control in Automated Highway Systems.  University of Southampton, Transportation Research Group http://www.trg.soton.ac.uk/rosetta/workareas/6a_ahs/ahs_synopsis.htm.  Accessed July 31, 2005

14.  Shladover, S., J. VanderWerf, Mark A. Miller, Natalia Kourjanskaia, and Hariharan Krishnan.  Development and Performance Evaluation of AVCSS Deployment Sequences to Advance from Today's Driving Environment to Full Automation.  California PATH Research Report, UCB-TS-PRR-2001-18 August, 2001

15.  PReVENT Home Page.  http://www.prevent-ip.org/en/home.htm.  Accessed July 31, 2006.  

16.  The Aerodynamic Performance of Platoons: Final Report, by Zabat, Stabile, Frascaroll, Browand, California PATH Research Report 10/1995, ISSN 1055-1425

17.  The Auto Channel review of 1994 Lumina van. http://www.theautochannel.com/vehicles/new/reviews/wk9348.html  Accessed November 18, 2005

 

Exhaust gas

EDITED OUT

Vehicle Types and Degrees of Automation

Traincars

Transitcar (unedited… this concept is a bone for bus manufacturers)

In order to provide a bus-like transit service, traincars may offer the ability to walk through to another connected vehicle.  This would allow a new type of operation wherein transitcars specialized for passenger transfer would make a series of local stops at low speed in driverless mode.  The last stop would be coordinated with a passing roadtrain bus…   last section of bus detaches in order to service one or more local stops. Passengers would walk to the rear of the roadtrain to enter this transfer vehicle (having mostly standing space) when they are planning to get off the roadtrain.

Such traincars must be able to autonomously join and leave passing roadtrains on a bus route.  They would automatically pick-up and deposit passengers at designated bus stops, coordinate with an approaching roadtrain, accelerate up to speed, merge and couple into the roadtrain.  Such transitcars would need to have the ability to service a series of local stops at low speeds in a driverless mode. 

 

Taxicar & Deliverycar  

More fully automated vehicles may rove though neighborhoods at low speeds in order to provide inexpensive taxi service.  This concept has been demonstrated by the Robocab and Cybercar projects.  REF   Similarly autonomous vehicles may also be adapted for delivery and pick-up of parcels.

Manualcar

It would also be possible to have less automated and therefore somewhat less expensive traincars which require a licensed driver.  These could simply offer an enhanced form of Adaptive Cruise Control (ACC++) that fully takes over longitudinal control during a critical period when joining or leaving a moving roadtrain.  These are times when the short following distance requires cooperative braking among vehicles much faster than afforded by human reaction times.

Manualcars would tend to be sold to individuals for mixed rural and urban use.  In increasingly rural areas, manualcars would be more conventional in terms of aerodynamic shape and powertrain.  However, these vehicles could still operate as part of a roadtrain.

Cartcar

A fourth type of relatively inexpensive traincar might also be offered for private use.  These would have transitcar-like ability to join roadtrains operating on the thoroughfares.  However, they could also be driven to and from the “bus stops” at strictly limited maximum speeds.  Like golf carts, these might be driven by unlicensed teens, elderly and people convicted of driving under the influence.

Locomotives

Locomotives provide power to mostly electric follower vehicles.  It must be driven highly competently and automation will play a substantial role in assisting the driver.  Leadership is a privilege subject to intensive monitoring and reporting. 

Leader vehicles may also operate as followers as in a tandem locomotive configuration. 

Summary of Major Roadtrain Vehicle Types

Roadtrain Car Types, Capabilities & Driver Qualifications

 

Docking

Merging

Door-to-door

All-electric

Licensed
Driver

Transitcar

auto

auto

no

yes

no

Taxicar

auto

auto

yes

 usually

no

Deliverycar

auto

auto

yes

 usually

no

Cartcar

auto

auto

yes

yes

maybe

Manualcar

auto

manual

yes

 usually

yes

Leader

auto

manual

yes

no

special

The table summarizes the preceding definitions.  Taxicars and deliverycars are most automated; they are able to rove neighborhoods providing D2D service.  Cartcars allow unlicensed drivers to ride D2D in vehicles that may be more readily available, fun and less expensive than taxicars.  Like transitcars, cartcars can merge into roadtrains without rider intervention.  However, transitcars are public vehicles that do not penetrate side streets.   Manualcars are automated only during the mandatory docking phase. 

New Roadtrain Vehicle Categories

Roadtrains will be heterogeneous with respect to vehicle type.  Besides the major distinctions between locomotive, hybrid-electric and all-electric vehicle, they will tend to be identified by external shape.  The categories will be similar to those available today and encompass such terms as heavy truck, bus, light truck, SUV, van, wagon, sedan and two-seater.  However, roadtrains will change the mix and should introduce some new design categories.

Retractable Tri-wheel

This traincar design provides a significant aerodynamic gain by eliminating the losses from one-half of the wheels.  This approach also greatly simplifies the steering of vehicles while following. 

Most traincars will be all-electric with golf cart like top speeds while operating independently.  Thus they will not experience the high lateral g-forces that cause tri-wheel cars to tip over.

A retractable 3rd wheel may also reduce vehicle weight if the front gear is lightly loaded and the vehicle subject to some (automatically assessed) weight and balance criteria. 

Single Occupancy Vehicles (SOVs)

Roll-over prevention requires squat vehicles with a low CG and wide wheelbase while aerodynamic efficiency suggests square and rounded cross-sections.  This will result in a poor trade-off for vehicles designed for single (or cozy dual) occupancy.  In turn, this tends to cause vehicles supporting at least tandem occupancy to be purchased even if rarely used other than by a sole driver.

Such vehicles can benefit from a design that allows them to form a rigid connection with one or more such adjacent vehicles.  This may be achieved by means of lubricated slots on one side with a matching tongue on the opposite side. 

Although the width of such vehicles need not be defined, their conjoined configurations may never exceed the width of their leader.  One natural sizing would allow up to 3 all-electric SOVs to fit within the width of typical locomotive.  Only one of the SOVs would use the centrally located couplings and any followers would also use its centerline coupling. 

Reversible

The ideal aerodynamic shape has a tapered tail as well as a nose.  Transitional cars, defined as cars where the cross-sectional area is larger at one end than the other could be designed to merge (automatically) facing backwards.  This would potentially double the number of scenic seats. 

Sleepers

When operating on divided, limited access roads over long distances, traincars should be able to operate with railroad levels of safety.  This will permit vehicles to be equipped with adjustable beds that can be made to lie flat or assume the shape of a bench for driving.  To comply with regulations on passenger restraint, these cars may also be equipped with airbags or nets on an energy absorbing reel. 

Compact all-electric batteries (which are also likely to be swappable to provide range to permit some off-road and rural uses), compact motors and even steer-by-wire will allow almost all of the follower vehicle volume to be used for comfort and cargo.  Combined with the need to optimize aerodynamics, the desire for interior space will favor traincars that are shaped like sections of an oblong box with rounded edges along their length.  This shape is apparent in the aerodynamically optimized intercity speed express train pictured here.

 
XXX need copyright approval

Replacement of heavily used buses can somewhat reduce the improvement measured in passengers per lane-hour because a unit length of roadtrain will be less densely packed than a full bus.  Highway lanes filled with such buses are reckoned to have a maximum capacity of approximately 30,000 passengers per hour whereas 10 foot long, four seat traincars might max out at about five times the current 2000 vehicle/hr limit or 20,000 passengers per hour based on 2.0 riders per car.  Both maximums may be increased by either platooning buses or, in traincars, increasing the fraction of seats used especially in somewhat larger (e.g. van-sized) cars.  In actual use, the gains in passenger throughput can be expected because of substitution for regular cars.

Beneficial Side-effect of Lightest Last Rule

The fact that the safest position in a roadtrain is at the end will serve as an inducement to choose cars with the least weight and lowest environmental impact.  Poor drivers may also end up in the lightest vehicles if other drivers decline to be put in a position where they may be lead by a substandard driver.

This paper begins an exploration of the technologies, procedures, laws and driver behaviors that could be used to assemble, operate and disperse roadtrains while operating in mixed traffic.  Combined with elements of aviation such as redundant systems and certifying roadtrain pilots with cooperative and professional attitudes, roadtrains show potential to reduce accidental losses more than tenfold.

SURFEIT OF LOCOMOTIVES this is likely to occur naturally but could be ensured with tax or other incentives).

For Traffic Engineers

Waiting traincars will need adequate space while holding for locomotives.  These will most often be created by eliminating thoroughfare parking spaces near the intersection.  Loss of such spaces should be compensated by reduced use of private cars.

Weaving and Running Naked (operating without a locomotive)

dating prior to mating AND A WHOLE RAFT OF ALLUSIONS

Roadtrains will sometimes have to slow down in order to pick-up traincars with depleted batteries that are unable to reach cruising speed.  The minimum signaling capability may involve emergency flashers and use of the intercom to communicate achievable speed.

Role of Road Condition and Speed in Stability Control

Roadtrains should never experience tire slippage.  This is especially true of the front wheel of the lead vehicle because any extraordinary yawing due to traction loss at speed will result in a serious situation.  The primary prevention strategy is speed control.  Roadway condition monitoring can play an important role in establishing speed limits with an adequate margin of safety. 

Physical Coupling Provides Additional Stability

Blowouts & Wheel Loss

Roadtrain cars must remain coupled in the event of a tire blowout.  Stability should be enhanced by mechanical coupling so that the affected traincar is virtually guaranteed to continue tracking the lead vehicle.  This can be also be ensured by using tires that run flat or need no inflation, and an appropriately low center of gravity. 

Loss of control due to total wheel separation must also be prevented.  This may be achieved by a combination of prevention and stabilization based on the connections to other traincars.  The former may be achieved by detecting vibration when securing lug nuts become loose, strain gauges to detect metal fatigue and visual inspection. 

Realizing the mutual stabilization attached traincars afford each other will establish a minimum strength for the coupling.  It also reinforces the constraint that heavy traincars may not follow lighter ones.

Sideswipes

Physical connection will tend to stabilize traincars experiencing collisions with other vehicles traveling in the same direction because the traincars ahead in the roadtrain will pull the affected traincar back into the center of the lane.  However, traincars behind may be adversely affected by secondary collisions.  These can usually be avoided by automatic decoupling and maximum application of brakes.  However, this would divide the platoon at a time when drivers in the following traincars are likely to inattentive and would require a Lane Keeping System.