the introduction of driverless cars will be hugely disruptive, a revolution comparable to the introduction of computers, the Internet and mobile phones. it will make roads safer, taxis cheaper, and radically alter the built environment. but, just as computers made millions of pool typists redundant, so driverless cars will make millions of taxi and truck drivers redundant. it’s a case of when not if, with countries reviewing and updating legislation to legalise driverless cars on public roads.
the benefits are so attractive, especially for city dwellers, that take-up will be rapid once regulatory hurdles are cleared and driverless cars are available to the public for use on all public roads:
- people of all ages, and physical and mental abilities will have access to personal transportation, owned, leased, shared or hired as needed.
- roads will be much safer.
- there will be no need for on-street parking, freeing up acres of space for other purposes.
- fully-electric cars will be viable in cities (where most cars are parked on the street, making charging batteries difficult).
- owning a car will not be necessary or cost-effective for low mileage travellers: driverless taxis will be cheaper than taxis now, especially for longer journeys, because fleets of driverless cars can reorganise themselves to be available where there is demand.
- road signs may be eliminated (vehicle priorities can be arbitrated wirelessly from roadside beacons or between vehicles).
millions of truck and taxi drivers will find themselves redundant in the space of a couple of decades, a potential societal disaster. in the US alone there are 3.5 million truck drivers. to begin with, truck drivers will, for public reassurance, be employed as in-cab supervisors; drivers will be employed at depots to marshal trucks until that process is also automated; and doorstep deliveries will require drivers until they just do the delivering.
though there are fewer taxi drivers (around quarter of a million in the US), their employment prospects are bleaker as there will be a much shorter transition period than for truck drivers before they become redundant.
it is imperative that governments prepare retraining programmes at the same time as considering how to regulate and adapt infrastructure to driverless cars.
a major reservation about driverless vehicles is their safety: surely robots cannot be entrusted with human lives? the truth is that people are not as good as we like to think: on average across the world, eighteen people out of every hundred thousand die in traffic accidents each year, or about 1.26 million in 2015. the vast majority of those deaths are due to human error. here is a comparison between the weaknesses that a human driver has compared with a computerised drive system:
|people||computerised drive system|
|inattention.||attention is always 100%.|
|misjudgement of speed or distance.||a combination of GPS, cameras, radar and wireless communication with roadside beacons and nearby vehicles will provide precise information on every vehicle, pedestrian and any other obstacle’s location, velocity, acceleration, and potential collision course.|
|awareness only within field of view.||cameras and radar will provide continuous 360° awareness.|
|poor night vision.||infrared cameras and radar will provide almost as good awareness at night as in the day.|
|tiredness.||computers don’t get tired.|
|slow reaction time (in the order of 1 second).||reaction time is in the order of milliseconds.|
|misunderstanding of other drivers’ intentions.||cars will constantly exchange data wirelessly to communicate their current state (velocity and acceleration) and planned manoeuvres.|
|mental overload leading to poor decisions.||computers can process millions of pieces of information every second.|
|inappropriate response to weather conditions.||systems will continually assess the road conditions and visibility and adjust the speed and manner of braking accordingly. traction control and antilock-brakes will be much more effective under computer control.|
|inappropriate evasive response to an imminent collision (e.g. swerving into the path of another vehicle, or braking while steering causing an uncontrolled spin).||when evasive action is required, the system will already have full awareness of where it is safest to move the vehicle to.|
|rash behaviour owing to rage, impatience, overconfidence or showing off.||computers do not feel emotions.|
|inadequate maintenance.||a driverless car will have more self-diagnostics, and can take itself to a mechanic without inconveniencing the owner.|
there will come a day, certainly by 2050, when people will wonder how we ever allowed humans, with relatively little training and assessment, to take complete charge of such lethal machines.
situational awareness is an important concept in the military: it involves building and continuously updating a mental map of yourself in your surrounding environment: where everyone is; what they’re doing; what they are planning to do; how they will react to any of a number of pre-examined events. it’s like a game of chess: knowing where all the pieces are on the board, the strategies at play, and the consequences of any move.
this is something that computers are particularly good at, even without possessing human intelligence. a regular desktop computer can use ‘brute force’ methodology, where thousands of scenarios are played out from the current state of the board, to beat almost any human chess player.
cars will gain situational awareness by multiple means:
- cameras sensitive to visible and infrared light (infrared images are useful in the dark and for locating people and animals, stationary or moving);
- radar (to locate and determine the speed and direction of other vehicles, cycles, pedestrians, animals and static obstacles);
- GPS signals (for navigation);
- wireless communication from roadside transmitters, in particular at junctions;
- wireless communication from nearby vehicles (say, within 200 metres), which will continuously broadcast their:
- velocity (speed and direction)
- acceleration/deceleration in all axes (forward-backward, left-right, up-down)
- current and planned maneouvres (e.g. turning or changing lanes)
- anticipated safe stopping distance
this information can be processed and mapped in real time, ensuring that all vehicles maintain a safe position on the road, at a safe distance from each other and from surrounding obstacles.
the HAL 9000 nightmare
understandably many people are reluctant to surrender control to a robot: it could malfunction in any number of nightmarish ways, killing its passengers. drive control systems will be complex, and inevitably they will contain bugs and will suffer failures. computerised visual recognition systems will make mistakes. the solution is in redundancy, using multiple systems to do similar jobs: inputs from all sensors can be cross-compared and compared with data received from nearby vehicles to create a completely accurate dynamic map of the vehicle’s surroundings. a computer system, similar to an airline’s ‘black box’, that is completely separate from the drive control system can monitor all sensory inputs and activities, checking for anomalies. where an anomaly is detected, it can initiate an emergency controlled stop and, at the same time, alert all vehicles nearby to take evasive action.
liability insurance that covers a vehicle’s driver is a legal requirement in almost all countries. in the case of an accident, each driver’s insurer pays out roughly in proportion to the degree of his/her fault. the fault might have been owing to a mechanical or electrical fault (an indicator light not working, or a brake failure), but the driver of that vehicle bears responsibility, even if s/he had taken all reasonable steps to maintain the vehicle.
for some years after their introduction, driverless cars will operate in a sophisticated cruise control mode with a manual override facility. their operation will depend on there being a responsible driver in the driving seat, ready and able to take control at a moment’s notice. the insurance model will not need to change while this is the case.
however people will learn to trust the autopilot and will cease to pay close attention to the road. drivers will still cause accidents by making mistakes when in control, or by not being attentive enough to take control when needed. so the drive systems will include increasingly sophisticated accident mitigation systems that do not rely upon, or even require, human intervention. it will then no longer make sense for the driver – if indeed there is a driver in the vehicle – to be responsible in the case of an accident.
it makes most sense for the vehicle owner to hold insurance cover for his/her vehicle, but it is not so obvious that s/he should be held responsible for any accident caused by his/her vehicle. indeed the finger would point toward the manufacturer and the software engineers behind the drive system as, it will be argued, it was their faulty or inadequate drive system that caused or failed to avoid the accident.
as long as victims in an accident are compensated by insurers in proportion to their need, it will not matter to them who or what was at fault; that will be something that the insurers will (as now) argue between themselves and with the manufacturers involved (or their insurers) where fault should be ascribed. for the manufacturers and their suppliers, the prospect of being on the receiving end of multi-million dollar law suits will provide a strong incentive to build highly robust and rigorously tested safety systems.
it will (as now) sometimes be impossible to ascribe fault, or fault may be ascribed to an uninsured vehicle owner. in these situations it is important that victims are still adequately compensated. this can be achieved either by legislating for a minimum insurance cover for all vehicle owners that covers injury for all vehicle occupants no matter where the fault lies. alternatively, countries might follow the example of New Zealand (c.f. Accident Compensation Corporation) and create a national accident compensation body, funded by taxation, that pays out entirely on the basis of need, regardless of fault.
currently maximum legal speeds are set nationally with some local modification. for the most part it is the class of road that determines the speed limit (motorway, trunk route, urban, etc.), taking no account of local variations, such as bends, concealed junctions, road width, etc. local modifications are usually only made in response to high accident rates, and are expensive to implement.
vehicle computers will determine the optimum speed of travel algorithmically, based on:
- what nearby vehicles are doing (which will be communicated wirelessly as well as sensed by radar);
- curvature of bends in the road;
- location of junctions and crossings;
- weather conditions;
- presence of pedestrians;
- width of carriageway;
- type and condition of road surface;
- engine efficiency.
Initially it is likely that governments will want to maintain absolute speed limits, but these will become irrelevant when it is computers that continuously assess what speed is safe in context.
road signs & traffic lights
road signs can be eliminated as they serve to communicate visually to human beings information that may be communicated wirelessly.
- directional signs will become redundant as vehicles will use digital maps and GPS receivers;
- speed limit signs will become redundant as vehicles set their speeds algorithmically (see speed limits above);
- traffic lights will be replaced by roadside beacons that communicate with approaching vehicles and arbitrate priorities responsively to ensure that traffic flows at optimal efficiency;
- give-way and other junction signage will become redundant as vehicles will negotiate priorities between themselves (just as people do now at give-way junctions and roundabouts).
car sharing schemes such as Zipcar will become hugely more attractive when cars are driverless. currently members cannot rent a car one-way: they must pay for every hour or day until they return the car to its place of origin, even if they are not driving the car for most of that time, and the car is not available to anyone else to use. a driverless car can deliver itself to another customer after each booking.
low-cost driverless taxis will compete much more keenly with public transport, especially local bus services. the risk is that city centres clog up with large numbers of vehicles carrying one or two people (and circulating empty between rides). local regulation will be essential in ensuring that this doesn’t happen. this will fall into two parts: limiting vehicle access to city centres, and promoting an integrated hierarchy of transport modes.
the only way to eliminate congestion in city centres is to limit the number of vehicles permitted access. permits should be granted on the basis of need and utility, giving priority to vehicles carrying multiple people (e.g. shared minibuses), people with impaired mobility, goods and equipment.
to ensure the most efficient use of space and the viability of public transport services, there must be seamless connections between all modes of transport: rail, bus, minibus, shared taxi, private taxi and cycle. this will ensure that shared transport and cycles predominate in city centres, and private taxis are found mainly in suburban and rural areas to provide the ‘last mile’ connection.
there will be no need for street parking: vehicles will only enter city streets to collect and set down passengers, and deliver and collect goods and equipment. only drop-off and pick-up points will be required every fifty metres or so. the released space may be repurposed for:
- cycle lanes
- cycle parking
- household refuse collection points
when not in use, vehicles will return to depots and high-density car parks, out-of-town or underground, where vehicles can be charged, serviced and valeted. the land occupied by existing city centre car parks is so valuable that it will be redeveloped for business and housing.
high capacity batteries (e.g. vanadium flow) at these car parks will smooth demand from the electricity grid. during the night, when grid demand is low, batteries of parked cars and the on-site vanadium flow battery will charge slowly. by morning, cars will be fully charged and ready to transport people to their places of work. when those cars return to the parking lot during the day, they will be charged from the vanadium flow battery (unless there is a surplus of solar or wind power available). cars will again be fully charged in time to collect people from their places of work. when those cars return, they will again be charged from the vanadium flow battery, until national demand for electricity falls to a level where everything may be charged again from the grid.
when grid supply is low, the vanadium flow batteries can feed back into the grid; and if critically low, car batteries could also be discharged to the grid. this smoothing of peak demand enables a higher proportion of power to be generated from solar and wind, and reduces the need to use gas-powered generators to meet peak demand.
refuse collection points
refuse collection will make use of driverless vehicles. in cities, where people now have to accommodate up to three bins (for compostable, recyclable, and land fill waste), it is likely that councils will gradually set up multiple collection points along streets where there were previously parking spaces. it may be that each collection point will be a car-sized refuse receptacle on wheels, possibly with a built-in compactor. the receptacle will probably have separate compartments for different categories of waste. once a compartment becomes full, the vehicle will signal for another refuse vehicle to come and replace it. if these refuse collection points were spaced about 50m apart, there would be one within 25 metres of any house. the short walk to the bins would create a new opportunity for neighbours to meet.