EAN DRIVERLESS CARS: Reducing Congestion on Expressways, but Increasing Congestion in City Centres There has been much recent talk of driverless cars, with high profile companies such as Google joining a number of leading car manufacturers in developing systems to enable driverless cars to take to the road. Richard Di Bona argues that whilst such cars may reduce congestion on segregated highways through increasing effective capacity, within city centres they may be more likely to increase congestion. (Sh AUGUST - SEPTEMBER 15 TatfelsaTech ww aicrratech com1 was not that tong ago. shen cciveress care were the stuff of scence ition. But given the relentless pace of development in computing systems, both in terms of increased speed fand reduced cost, as well asin visual recognition, such drivers cars are naw deemed largely feasible Indeed, there has been everarowing press coverage af advances in automating cas, such that they can navigate and dive Without any interaction from humans. This elther for driving with automatic collision avoidance on highways or aven to contra a full journey, with humans ‘merely stipulating the trip's destination and perhaps aso the time to make the trip. There has also. been substantial commentary on how autonomous cars rmght improve road safety, road capacity and road ervironments m general. Whist there are indeed grounds for optimism, there remain chalenges tobe addressed, including some possible negative futomes, which appear tohaveatvacted less media coverage. This article is thus intended to provide some balance How long have driverless cars been in development? Whilst the Google™ rveress cor is perhaps the most publicised, i is by na means the only, nor indees the fst drvalss car-under development fand_ testing. The first autonomous ‘demonstation vehicles appeared in the 1980's, though these were not used on condnary roads wath other road trafic It ‘was withthe advent of cheaper and faster Sensors and computers in recent years that autonomous cars. potentaly capable of ‘aig with other traffic became 3 ralty Inthe rite tates, the National Highway Traffic Safety Adminstration (NHTSA) proposed a formal dassicaton system to ‘eserbe the level of wehiceautonomy/ sl ring capably, a Fellows + Level 0: the river completely contol the vehicle aa imes + Level 1: ingvidual controls are automated, suchas automaticbraking + Level 2: atleast two conto can be automated together, such as acantiee ‘ruse conto ad lane keeping ‘won tatcintctch com * Level 3: the driver can leave all functions to the ear under certain situations, but the driver remains Comhand, withthe cer sensing when control should be returned to the iver + Level 4: the vehicle can perform all navigation and safety functions on ‘own, induding parking: the vehicle can operate without a driver being present Progress with driveres cars now makes Level 4 autonomy realsable. The question then becomes what ace the Hely pacts ‘of such fll autonomy for vehicles? As will be argued, the impacts are likely to vary between segregated highways and mixed traffic urban areas, Increasing Capacity on Highways Considering the potential impact of rere carson highways as cppesed to tyes, 12 on re thout pedestrians potently crossing, Kerbside parking or frequent junctions [potentially requiring vehicles 10 tum against anposing taf, rues cars have the potential not only to reduce accidents, but ako to increase read capacities and decrease congestion. Driveress cars would have cellsion avoidance algorithms, which could perform braking or evasive manaewwres ‘more quickly than humans could react However, overand-above this the potentl fr plotoning beheviou Under traditional, human-controlled sein, the capacity a a highway in terms ‘of cars per hour or eats pe ane pe hour can be described using the Fundamental Diagram of Trafic How, as illustrated below. Figure 1) shows the national relationship between trafic speed and the density of vehicles, reflecting the sareater typical spacing between cars as, speed increases, The consequence of this relationship on trafic flow versus trafic density is shown in Figure i) and onthe resultant relationship between speed and traffic fw in Figure 1, Figure 1: The Fundamental Diagram of Trafic Flow (for Humar-Contraled Vehicles} ese Density versus Speed Figure 1 Speed versus Taff Density ‘ena vers Flow i iT \ Figure fi Trafic lw versus Trafic Density Flow versus Speed se i> Figure Wal): Speed versus Tac Fow What the Fundamental Diagram of Tetfic Flow shows i that above a certain speed, the capacity ofa highway stats to cine. As such, there are two ways 10 Increase a highway’ capacity, either + tempt to manage tatfie speeds to always be aroun that level which rmaxinises capac, of + Find seme method to safely allow ‘echices to folow one another more dosly at higher speeds Drivetess cars, which ore not only ‘continuously sensing their surroundings, but which moreover communicate with fone another may enable the second of the above possbiltes. Under computer >» ‘TralfcintraToch AUGUST SEPTEMBER 15.65TSN contol ensuring thatthe speeds of vehicles traveling ina platoon or convoy are bept virtually identical, t becomes possible for ‘ats to be spaced far mare dosely than would be deemed safe under human conta. The effects ofthis on attainable traffic speeds ard highway capacity are showin in Figure 2, wth equivalent values for human cont shown nEghter shades Figure 2: How Drverless Cars Might Change The Fundamental Disgram of Trafic Flow (Odveress Vehicles shown Bold: Human Controlied Vehicles shown fain Density versus Speed ‘fc ent Figure 20): Speed versus Trafic Density wn = — aon Figute 2) Speed versus Tati Flow (56 AUGUST SEPTEMBER 15 TatfeloaTech Figure2 therefore showsjust how much capacities and speeds might increase with {@ deployment of driveless vehicles able te communicate and coordinate wth one another Figure 2 assumes a lane or lane(s) of traffic given over to the exclusive use of driverloss vehicies. In mixed trafic, ie, a combination of drivercontolled vehicles end computer controled vehicles, the scope for platooning is key to be substantial reduced, Nevertheless, this stil shows the potential for eriverless vehicles to boost Capacity and speeds on highways in an interurban contest, efor expressvays and the ike, But these benefits do not necessarily apply to urban environments! Some commentators have claimed that similar Benefits could arise fom the deployment of driverles ears in uroan environments, ie, with respect to speeds and capacities. However, thi ignores the realty thatthe speed of divers cars in ‘aed traffic would be constrained bythe speeds and behaviout of ether road users, such as humencontiolled cars, other vehicles and a0 pedestrans cessing the read. Alo, in mary urban environments, the criteal determinants of road network capacities and journey times are likely to be delays at junctions mere than jourey speeds belwueen junctions Inf, there area numberof oherissues with cveress cars in urban environments ‘which lea need ta be considered Potentially doubling vehicle kilometres travelle One of the benefits of criverless cars touted is that they could either drop their lower at their workplace then 90 and park themselves, or coud chive home, 50 3510 reduce the spacetae in city centies voted to car parking, Whist reducing landiake for parking could be beneficial such clams ignore one consequence of this: namely tha cars would end up rng twice as fara day. as stead of one round: tip fom home toa city centre car park and then back agany they would nave {rom home to the owners’ workplace then back tohome and then repeat this journey when itis time to collet the ovine. This would increase pollton (unless all srveress cars were electrically powered and all power generation was from polation, renewable sources) Moreover, it would roughly double traffic levels on city streets, hence Hkely resulting in increases congestion, Problems for traffic control systems In fac, congestion levels could more than double. At present, traffic signals take advantage of what is termed “tidal lowe” Put simply, in the moing peak there i more traic traveling tothe cty centre than inthe reverse direction; whist in the evening peak ther is usualy more trafic Fewing cut of the city centre ‘As such, optimised trafic signals tend to ive more green time to those movernents withthe greatest flows. If driveress cars were to retur home after dropping ther ‘ner at work and then dive back into the sity to collect them, then not only would overall traffic levels increase, but fen more ential, traffic flow would no longer have the same level af ily: Consequently, average delays at junctions would increase for allvehiles. nd where ‘ily networks are close to capac (as = often the cas) this also razes the danger of trafic queues at intersections blocking back to ther other junctions upstream. There becomes a larger danger of gridlock, Increased Vehicle Circulation ‘A further problem is likely to atse, whether cars return home during the day or whether they park themselves doyntoun and then come to cla ther ‘wes when their overs wish to return home, Uniessarduntl the frecoursan fr other vehicular circulation facilities of boulings ae overhauled there i unikely tobe sufficient capacity for vehicles to wait to collet ther anes. Such remodeling of blcings facies tkaly to be quite timeconsuming and cally, even where retrofiting is possible. And more tkely, the space requirements for this would ey undo a large preparton of mooted ow raticinataen comlortab savings from reducing downtown pking spaces consequently, itis ighly key that vehicles would need to ciculate in the verity of ther ones’ workplaces whist _anaitng ther one's aval fer colecton Such vebicuar creation would of curse further increase trafic congestion levels! ‘A Possible Paradox of Increased Vehicle Col ("Car Crashes") Whit colision avoidance mechanisms bulk int drveress ears would make them ser, there is so a paradoncal danger of such vehicles increasing overal accent fates on urban road networks ns ‘Auman driver woud quicy realise that dverless cars are programmed to avoid ‘llsions. As such, 3. human controled vehicle 1s key (0 be able to squeeze computercontaled cars mace effective than cars driven by other humans, for instance when wishing 1 change lanes or totum onto a road. The ccmputer controls in the other car would piovitse calison ‘avoidance and #0 vould cede spac 10 more aggressively (human) civen cars, The problem is this: @ human driver becoming conditioned to being able to ‘squeeze infront of computer contralled vehicles may well assume that a vehicle is computercortelled, only to discover too late that i was under human contol ‘nd the human diving the other car, with “ower reaction times than te computer ‘and likly aso having a more aggressive mindset than a “robot” car, may wel end upina collion. After al, there ceulé stil be someone inthe drive's seat when 2 ‘aris under computer contol, sit woud not be readily apparent which cars are being acively driven by their occupant and whieh are being iven by computer Pedestrian safety and a collapse in traffic speeds and road capacities Another related problem pertains 10 the interaction between pedestrians and verles ears. Drivers car would need tobe programmed to avoid calsons with pedestrans, inluding any who might Jump off of the pavement/ sidewalk ‘won tatcintctch com and inte trafic with ile oF ne warning Consequently, divrless cars wil need to respond instantaneously to any such instances of pedesvians stepping into the road This raises a number of possible problems. Firstly, pedestians are Tkely 0 fuck realize that driveress cars would allow them to coss the road. Pedestians ‘are likely 10 take advantage of this; 50 Urban traffic speeds could collapse as pedestrians make a habit of stepping into traffic more often than they currently do, As such, traffic may become very frequent interupted, thus resting in {ar lower effective capacties on 10365 However, bearing in mind that for some time to come after the introduction ff driveriass cars, that they would be sharing the road with human-contallee Vehicles, there i an inereaced risk of 2 human controled vehicle rearending 2 computercontalled vehicle. which comes to a sudden hat to avoid hing 1 pedestrian. On thi last pint, it might be argued that the human controlled cat could have some colision avoidance technologies; however, thi presuppares that al vehicles would quickly switch to such teemooaies It could be argued that the presence cof human controled cas, and the danger of jaywalking pedestrians being sued when a human contralled car rearends 2 diiveess car due to the jaywalking pedestrian stepping into the road, would tend to mitigate against such accidents happening. However, in many lursdctions ths could require ® change to how insurance lili i apportioned, possibiy itself requiring 2 change in legislation. Indeed, it haz been argued that clweless cars ae likely only 10 widely feature in society, nat only after a raft Of legislation has been revised, But aso folowing 2 numberof legal es cases Cities for Pedestrians or for Cars There is litle doubt that over recent decades the trend in most cities has been to prostise the motorcar, in terms ese Of allocating not only road space, but a0 patking space; often resulting in pedestrian beng forced onto feotbridges or underpasses. However, in some cities, primarily though not exclushely in Europe, & countersrend is now emerging. Mary cites ee now looking to increase the pedestrianisation of areas, so as to improve the ufoan envionment. Also in many cites, both in Europe but moreover also 1 much of the rest of he World, including india, these i an inereased emphasis on promoting pubic ransport; this often combined with ‘demand management measures for car usage. Given such trends, one also has to |queston whether rot cars, even under computer control, would cominue to be accepted in great numbers city conte Bear in mind that ould a time come when the overselingly majority of cars axe fully computer controlled, then so a= {o avoid issues with pedestians stepping ut into the road at wil, there would need tobe a masse stepup in legisaton and enforcement against jaywalking, teen an minor roads. In many cies, this \Would recuie a social hit to accept such prohibitions on freedom of pedestrians to ‘oss roads; and in many places such a Socal shift woud be greater than that requited to accapt divetess casper se Based on the above, | would contend that diverless, or fully computer: controlled cars + Are tkely to have a substantial role im increasing the safety and capacity Of seoregated highways, such as expressways and motorways: but, ‘xe uniely tohave a substantial and ‘widespread role within city centres (excepting any segregated highways fassing tough urban areas) (Pickard Di Bona is 3 Hang Kong based lransport consultant sith over 20 years of consultancy experience gained trom projects in over 30 countries. He has Undertaken 3 wide range of transport planning, demand forecasting, economic appraisal and transport policy projects). eco ‘TralfcintraToch AUGUST SEPTEMBER NS. 67