Canberra Autonomous Car Simulation

Public transport in Canberra

The city of Canberra presents a problem for transport planners. With both a low population (less than 400,000) and low population density, Canberra residents are dependent on the car.

Patronage of the public transport bus system, ACTION, is low and has resisted all recent attempts at being raised [ABS Census data 2011, BITRE Information Sheet 33: Recent bus transport statistics from 2009, Canberra Times: Adult bus passenger numbers stagnating, July 2012]. As reported in the Territory and Municipal Services Annual Report, 2013-14, customer satisfaction is just 65%, only 71% of services run on-time (within a 5 minute tolerance), and fares cover barely 15% of operating costs, leading to a community subsidy of around $120M per year. Patronage declined even after the introduction of $12/day pay-parking to 9000 formerly free car spaces in Parkes and Russell in October 2014. Each passenger boarding costs ACTION $7.96 and with an estimated 1.3 boardings on average being required to complete a one-way journey, each such journey costs ACTION around $10.35 and the rate-payer almost $8.80.

The problem isn't so much ACTION; with a decentralised and low density urban plan, Canberra was designed for the car. Attempting to provide any public transport system which balances costs with service level is a thankless task. A service with sufficient coverage and frequency to provide an alternative to private cars is too expensive. ¹

The government is keen to remove Canberra's dependence on the car. Parking fees have been introduced and steadily increased². ACTION's budget and services have been expanded. Planning has begun for a single-route light-rail system to connect Gungahlin and the city centre with an initial budget of around $800M. The goals behind these initiatives are laudable and deserving of serious commitment: combating increasing traffic congestion and pollution, reducing land devoted to car parks and revitalising the urban landscape, reducing transport costs, providing greater mobility, health and safety for all citizens, and reducing disadvantage caused by lack of access to efficient and affordable transportation.

However, no incarnation of the current bus network nor even an extensive light-rail system can be imagined which meets these goals, short of "starting again": abandoning the 100 sprawling, far-flung suburbs of the "bush capital" for a high-density new city. Hence, both bus and light-rail are distractions from meeting the city's transport goals, but are being pursued only because no better alternatives are perceived.

Perversely, Canberra's planners continue to increase the requirement for transport by reversing previous policies to decentralise employment and services into town and group centres, preferring instead to concentrate employment growth in Civic and Brindabella Park, both remote from areas of residental land development and population growth, and reached by increasingly congested roads.

We assert that the widespread awareness of a much better transport alternative is imminent, and the dramatic improvements it offers to the efficiency, cost and universal accessibility of urban transportation will make its incremental adoption inevitable within 6-12 years.

Public transport is a public good. Whether by providing access to education and health services, employment, or just facilitating independence and enjoyment of life, the benefits of increased mobility accrue to the traveller, their families and the community. Inadequate transport options diminish us all. Expensive public transport creates funding tensions which ultimately exacerbates social division.

Autonomous cars

The promise of autonomous or "self driving" cars have remained unfulfilled since the 1930's. However, developments in the past decade have finally made the commercial availability of mass-produced autonomous cars a strong likelihood within the next 10 years.

The 2012 KPMG white paper Self-driving cars: The next revolution presented a compelling case for the economic and social advantages and opportunities offered by autonomous cars and for a "mobility on demand" model based on a pool of autonomous cars. Their follow-up report from October 2013, Self driving cars: Are we ready? observed:

A year later the momentum around self-driving vehicles is astonishing. In some ways, the industry is moving even faster than we predicted. Rarely does a day go by without another announcement about a new technological breakthrough or a new joint venture. Traditional automotive manufacturers are teaming up with high tech companies; innovative start-ups are seeking and finding investors. The landscape is shifting before our eyes.
 
... The growth in self-driving mobility on demand services could mark the end of the two-car family ...
 
... we believe the market opportunities for self-driving vehicles and technologies are enormous, and innovative companies will continue to drive the technology forward.

The Accenture report The New Road to the Future: Realising the Benefits of Autonomous Vehicles in Australia from November 2014 reinforces this view:

Autonomous vehicles represent much more than a technology revolution; they require a complete transformation across the mobility ecosystem. There will be enormous disruption to established business models across multiple industries and the time is now to respond and position for the opportunity..
 
Australia offers the ideal market to test, produce and refine autonomous vehicles. Autonomous vehicles are already being tested on highways and roads, and commercial production has been flagged as three to five years away by most of the major car manufacturers.

Among the major players announcing intentions to deliver fully autonomous (also known as "Level 4") cars are:

• Nissan: Nissan pledges affordable self-driving car models by 2020, Press release, 17 Aug 2013
• Tesla: Tesla CEO Musk Sees Fully Autonomous Car Ready in Five or Six Years, Wall Street Journal, 17 Sep 2014. Musk expects Tesla cars to offer "90% autopilot" for highway driving by the end of 2015 (but that last 10% may be the hardest). In December 2015, Musk updated his forecast: "We're going to end up with complete autonomy, and I think we will have complete autonomy in approximately two years."
• GM: Emerging Technology: Driving Safety, Efficiency and Independence: "we expect semi-autonomous vehicles to be available to customers before the end of this decade and the technology for fully autonomous vehicles capable of navigating the roadways ready during the next decade".
• Ford: executive chairman, Bill Ford 2025+: "Arrival of smart vehicles capable of fully autonomous navigation", CEO Mark Fields (in Jan 2015): "We believe in the industry that there will be a fully autonomous vehicle, probably within the next five years". In August 2016, Ford announced their intention to deliver a "high-volume, fully autonomous vehicle for ride sharing in 2021". This vehicle will have neither pedals nor steering wheel.
• Audi: head of product communications, Stefan Moser, claims that Audi will be ready with fully autonomous vehicle technology in about two years, adding "When we show you something you can be sure it will be in a car in the next one and a half years. It’s important to show things that will come, not just dreams.”
• Benz: Dr Herbert Kohler, Head of Corporate Research and Sustainability and Chief Environmental Officer: "We are convinced that autonomous driving will be a central factor on the way to comfortable, accident-free driving"
• Google: Google driverless cars will arrive in 2017. Whilst not a manufacturer, Google is expected to partner with car manufacturers to commercialise its autonomous driving technology.
• BMW: with partners Intel and Mobileye, BWM plan to bring full autonomy to production cars by 2021. "This establishes the opportunity for self-driving fleets by 2021 and lays the foundation for entirely new business models in a connected, mobile world".

The advantages, disadvantages and economic implications of autonomous cars have been widely discussed. For dense cities with well-developed mass transit, they may have relatively little to offer. However, for cities such as Canberra they provide a method of better utilising the existing infrastructure of an extensive road network in a manner which meets the goals the government has identified:

• reducing congestion by vehicle sharing and automatically optimised traffic flow
• reducing pollution by operational and financial synergies with "zero tailgate emission" electric vehicles
• reducing land used for parking: autonomous vehicles can be shared and do not need to remain idle in the office or shopping-centre car-park
• reducing transport costs: as this simulation demonstrates, on-demand, 24x7 door-to-door personal transportation can be achieved much cheaper than through personal car ownership or conventional public transport options
• improving safety, mobility and social inclusion by providing very economical on-demand, 24x7 door-to-door transportation for all

The Morgan Stanley research report on Tesla Motors (Feb 2014) included a detailed analysis of the economic benefits that they think will drive the development and adoption of fully autonomous cars in the next 5-10 years. They assert that when almost 100% of the cars on the road are autonomous, a "base" estimate can be made for annual savings to the US economy of:

• Fuel savings: $US158 billion
• Accident savings: $US563 billion
• Productivity gains: $US422 billion (freeing drivers to do whatever they want)
• Congestion savings: $US149 billion

These total to $US1.3 trillion, or about 8% of US GDP. Morgan Stanley's "bull" and "bear" estimates for these savings are $US2.2 trillion and $US0.7 trillion respectively.

The Eno Center for Transportation's report, Preparing a Nation for Autonomous Vehicles (October 2013) estimates the annual economic benefits of autonomous vehicles in the USA as a more conservative $US447 billion at 90% uptake.

The Eno Center analysed the US Department of Transportation's National Motor Vehicle Crash Causation Survey, Report to Congress (July 2008) to estimate that 93% of motor vehicle accidents had human cause as the primary factor, and quotes Hayes' analysis that autonomous vehicles could reduce motor vehicle fatality rates per person-mile travelled to a level approaching "those seen in aviation and rail, about 1% of current rates". Human beings make terrible drivers³.

The annual cost of road crashes in Australia is $27 billion according to the Department of Infrastructure and Regional Development, or roughly $1150 per person per year. The Australian Institute of Health and Welfare's Injury among young Australians report from 2008 showed that transport accidents were the leading cause of death and injury of young Australians (aged 12-24 years) in 2005.

By removing human fatigue, distraction, inebriation and poor judgement as factors and by removing risks caused by speeding and running red lights, autonomous cars have the potential to significantly reduce injuries and death caused by traffic accidents.

Recognising that "the question is no longer if but rather when autonomous vehicles will appear on our streets and highways", the transport and logistics giant DHL is planning to use them to deliver innovative long haul and last-mile transport operations.

Recognising the economomic and societal benefits offered by autonomous cars, the governments of Singapore and the United Kingdom are building national expertise in the design and operation of new and more efficient and effective approaches to transport based on their use.

Electric vehicles

For all their "green" advantages, electric vehicles are still expensive. Whilst that is likely to change as battery technology improves and electric vehicles are mass-produced, like conventional cars, they represent an under-utilised asset. Most cars spend the vast majority of the day idle, and the capital cost of the car is defrayed over a small fraction of the potential travel utility it could be providing.

A great advantage of electric cars is their running costs: just a few cents per kilometre. Hence, the more kilometres travelled per day, the more the economic advantages of electric cars accumulate. But most current electric vehicles are hampered by comparatively short range and long recharging times which makes utilizing them continuously inconvenient or impossible. Managing them as interchangeable "transportation units" in a shareable fleet rather than as single vehicles is the key.

One approach is a small-scale (neighbourhood) car sharing scheme, but these are inconvenient: when you want a car, you want it in your driveway or at the office or outside the supermarket where you are standing. You don't want to negotiate with your co-owners or walk 400m to find it.

A fleet of autonomous cars provides a mechanism to implement a city-wide car sharing scheme at a scale which means a car is always available to meet your needs at a fraction of the per-kilometre cost that you'd incur if you owned it.

Aside from their contributions to greenhouse gases, the exhaust from conventional cars has a large negative effect on human health. The Bureau of Transport and Regional Economics working paper Health impacts of transport emissions in Australia: Economic costs, estimated that pollution from motor vehicles (cars, buses, trucks, motor-cycles, light commercial vehicles) was responsible for between 900 and 4500 cases of morbidity and between 900 and 2000 early deaths each year, with an annual economic cost of between $1.5 billion and $3.8 billion. This study estimated that by 2020, as a percentage of all motor vehicle emissions, cars would be responsible for 76% of carbon monoxide, 58% of nitrogen oxide, 67% of sulphur dioxide and 58% of particulate matter of less than 10 microns (PM₁₀).

Electric vehicles have zero tailpipe emissions, and when powered by electricity generated from renewable resources, remove this burden of pollutants from the environment in general, and human health in particular.

The motivation for this simulation is to explore the practicalities of car sharing using autonomous electric vehicles. Specifically, we wanted to model the public and private transport needs of Canberra and then test the ability of a fleet of autonomous electric vehicles to meet those needs.

As well as the "macro" transport goals listed above (congestion, pollution, land use...), we frame our consideration with the transport goals of typical citizens, such as these:

• Mary lives in Farrer and works in Barton. Each weekday morning she drives her 3 year old son to child-care in Narrabundah, drives to work and collects him after work. She often likes to visit the supermarket at Torrens on the way home because it stocks some of her favourite items. Today she has a dental appointment at lunch-time in Braddon - where will she park? Next Saturday she wants to watch her nephews play soccer, but she has to get to Amaroo by 8:30am for the kick-off.
• Joe lives in Scullin and works stocking shelves in a supermarket in Weston three nights a week. His shift starts at 8pm and ends at 4am. He is also studying part-time at the Bruce CIT, and is usually in a rush to leave Bruce and get to Weston to start his shift.
• Edith is 85 and feels she really shouldn't be driving, yet is not willing to give up her license and hence her independence. Her arthritis makes it impossible for her to check the air in her car's tyres, and to save money, she hasn't had the car serviced for three years. She drives her husband to medical appointments each fortnight and ferries home the bulky supplies his condition requires. About once a month she drives to her son's house and baby-sits his children, getting home after 11pm.
• Blake is out late again with his work colleagues. It's 2am and he's in a club in Kippax, wondering whether to bother even trying to get a cab to get back to Downer, or whether he should just risk getting a lift with Jed, who really shouldn't be driving either. But how will he collect his car from Kippax tomorrow?
• Jamie lives in Kambah, works at Campbell Park and is studying part-time at ANU. She often needs to work late, and also spends long nights at the Chemistry lab at the university, but has to leave before the last bus. She often feels anxious about her safety waiting at the bus-stop and walking home in the dark, and she's wondering whether she should give up uni or buy a car.
• Henry is in year 12. He doesn't yet have his driving license, and relies on his mum to take him to work his weekend shift at the pizza restaurant. The nearest library is 10km from home, but he enjoys studying there, so he puts up with the travel. His college day starts at 10am on 3 days of the week but his only option for getting to school is catching the bus at 7:30am each day. Henry's parents are separated, and although he'd like to visit his dad more often, the return bus journey takes nearly 3 hours on the weekend.
• Jayne from Higgins is a single mother of two young children. She suffers from a spinal condition which makes walking difficult and slow. She is unemployed, receives rent assistance and cannot afford to run a car. The nearest bus stop is 450m from her house. She started a hospitality course at the Reid CIT but withdrew because just getting there involved either three buses with a connection she inevitably just missed, or two buses into Civic then a 800m walk. Just getting the kids to doctor's appointments and the shopping done is hard enough.
• Hoang from Palmerston works in the Parliamentary Triangle. Before pay-parking was introduced, his daily commute took 20-25 minutes each way if he left home before 7:20am or after 8:45am and left work before 4:30pm or after 6:30pm. To save money, he now catches the bus, which takes 55 minutes in the morning and 50 minutes to get home if it runs on time. Occasionally, he has to work late but then has to be careful when he leaves to avoid a long journey home. Although Hoang saves money on fuel and parking, he now has 5 hours less time each week with his young family and can no longer help with shopping on the way home.

Most people would recognise these scenarios: most of us live busy and complex lives, juggling many responsibilities. Simplistic transport options rarely meet our requirements, making us either reliant on our cars or at risk of social exclusion and disadvantage.

No-one really wants a light-rail system, or a bus system, or even a car, autonomous, electric or otherwise. What they want is to be able to get to work, school or uni, get home safely from a party at 2am, visit the doctor, pick-up the kids from child-care, and on the weekend, take them to the soccer in the boondocks of that new suburb. They want a way to travel safely, cheaply and quickly from door to door, whenever the need arises.

As Simon Corbell, ACT's Minister for the Environment and Sustainable Development said in his introduction to Transport for Canberra - Transport for a sustainable city, 2012–2031 people want "... a transport system that puts people first ... [that] will make our city a better place to live, work and do business, and a more accessible place where it is easy for everyone to get around."

Urban and transport planning occurs in a context of multiple decades. Transport infrastructure is expensive with a long life. Urban plans, such as Canberra's, and the associated suburbs, housing, workplaces and leisure and shopping facilities have an even longer life.

Whilst not certain, the likelihood of autonomous vehicles becoming a reality in the next 6-12 years, is very high. Many automobile manufacturers, experienced commentators and industry experts would all have to be wrong for this not to happen. Yet currently, Canberra citizens are debating the merits of an $800M investment in a single light-rail line, which, if it is to expand to become a viable part of Canberra's transport infrastructure, will be just the first stage of an investment which would require over $10 Billion. Whilst no-one can feel confident that such a system will meet the needs of citizens even at such a cost, effective alternatives are not being debated. This simulation aims to test the feasibility of an alternative.

This simulation arose from a spreadsheet created by Warwick Cathro and Kent Fitch as part of Warwick's submission on the ACT Government's Low Emission Vehicle Strategy discussion paper. A spreadsheet can only go so far in confidently modelling scenarios as detailed as a city-wide autonomous car infrastructure, but a simulation enables assumptions to be reified, tested and corrected.

This simulation is the work of transport amateurs. Whilst it consciously attempts to err on the conservative side, it doubtless contains inaccurate assumptions, misunderstandings and plain old bugs. It is being made public in the hope of reducing these errors.

The simulation is written in javascript. All the data it uses is contained in the javascript it loads: there is no data in any database.

The following table shows typical outcomes when the simulation is run with the six pre-configured uptake scenarios which represent an increasing passenger load on a simulated fleet, rising from a low 45,000 journeys per day (equivalent to the journeys currently provided by ACTION) to 1.1 million journeys per day:

Each simulation run will produce different results because there is a random component to both the number of journeys starting each minute and the origin and destination of each journey: the results shown above are typical.

Apart from number of journeys, cars and chargers which are specific to each uptake scenario, these simulations were run with some common assumptions:

• Cars: Each car costs $40,000 commissioned, residual value: $0, useful life: 36 months, financed at 10%, maintenance: 2.5 cents/km, annual registration, insurance, admin, comms: $2,000, theoretical max range: 240km (but cars only charged to 80% of range and recharged at 25% of range), travelling 6km/kWHr and with a set of complete spares sized at 5% of fleet.
• Charging: Each charging station costs: $15,000 installed, residual value: $0, useful life: 120 months, financed at 10%, annual rent & maintenance: $3,000, power delivery rate: 75kW, cost of electricity per kWH: $0.20.
• Fares: Peak period flag-fall: $0.45 and rate per km: $0.25, cost of average 13.4km trip: $3.79. Off peak period flag-fall: $0.20 and rate per km: $0.20, cost of average 13.4km trip: $2.87.
• Miscellaneous: Fixed annual system cost: $1,000,000.

You are encouraged to configure the simulation to change these assumption and rerun the simulation for yourself. More information is available about these assumptions and how the model uses them. For example, a simulation based on some reasonable assumptions of using the Smart Electric Drive two-seater with a small range produces considerably more favourable results.

A graphical representation of the sensitivity of the operating surplus and wait times for the "Very high" uptake scenario to various parameters is available.

Some observations on these results:

1. An autonomous car fleet can provide the same number of journeys as the current ACTION network at less than half the cost

   The simulation of the ACTION load of 45,000 journeys per day has annual costs (operating and capital) of approximately $58M, income of approximately $46M, for an annual loss of $12M. Hence, all travel could be made free for a cost to rate-payers of $58M, less than half of the rate-payer subsidy to ACTION.

   Additionally, travellers would enjoy an 24x7, door-to-door, on-demand service, with 96% of journeys starting within 1 minute of being requested and over 99.5% of journeys starting within 5 minutes, even during peak periods. Tailpipe emissions would be reduced from an estimated 92 grams of CO₂-e per passenger km (see below) to zero.

   However, traffic congestion would increase, particularly during peak periods on major roads (due to major-route buses carrying up to 100 passengers being replaced by autonomous vehicles carrying between 1 and 4 passengers), making this an undesirable scenario. [Note that congestion can be reduced in this scenario by increasing wait times slightly by waiting for an extra minute before leaving a location if it is likely another request to a nearby destination will be received for that location. The poorer service levels, whilst still vastly superior to a bus service, may be acceptable, and would have the side-effect of reducing the required fleet size and hence costs.]

2. An autonomous car fleet can service a very high load replacing most private cars journeys and generating a large annual surplus

   The simulation of a very high load of 750,000 journeys per day generates an annual surplus of approximately $78M. This surplus could be used to provide 75,000 free off-peak journeys per day, or provide funds for other community needs, or some combination of goals.

   Over 98% of journeys start within 1 minute of being requested and 99.9% of journeys start within 2 minutes, even during peak periods.

   Traffic congestion is dramatically decreased, particularly during peak periods on major roads. For example, the average occupancy of cars arriving is Civic and Parkes is 2 passengers, compared to an estimate of 1.13 for current journeys to work.

3. An autonomous car fleet can service Canberra's future traffic loads using the existing road infrastructure and make a substantial contribution to government income whilst reducing transport costs for citizens

   The simulation of a "future" load of 1.1 million journeys per day generates an annual surplus of approximately $167M. The average occupancy of cars arriving is Civic and Parkes increases to over 2.1 passengers, so traffic increases at a slower rate than passenger journeys. Furthermore, at such high levels, forecasters predict that traffic scheduling algorithms that take advantage of vehicle-to-vehicle communication and coordination will smooth traffic flow and decrease travel times.

4. At lower uptake levels, a higher car-to-journeys ratio is required to achieve acceptably short wait times

   With a small travelling population, requests to start a trip are more "bursty" and hard to predict. Hence more cars need to be deployed across the city just in case a burst of trip requests are received at a location.

   To see why, imagine an average of just 1 trip per minute leaves a suburb and trips for 3 consecutive minutes are delayed by the traveller. If cars were allocated in anticipation of their arrival, all 3, or 100% of the allocated cars will be idle (hence wasted). Conversely, if these 2 of these requests were brought forward to arrive in the same minute as the first request, then 3 cars would be required simultaneously, because it is unlikely that this small number of requests are sharing a common destination which would allow a single car to be used.

   Contrast this to a large travelling population, where 10 requests leave a suburb on average. In this case, it is relatively "cheap" to over-allocate 1 or 2 cars per minute (10% or 20% over-allocation), it is relatively unlikely that many of the trips will be delayed or brought-forward, and it is more likely that some trips will share a common destination, allowing a smaller number of cars to be used.

   Over-allocation of cars creates "waste", yet under-allocation creates long wait times, and the simulations with small journey volumes struggle more to balance these undesirable outcomes. The more unpredictable or "bursty" the requests, the larger the effect.

5. Given the same levels of service, higher uptake levels generate a higher surplus

   The size of the fleet required grows slower than the increase in the journeys it needs to service because less over-provisioning is needed to cope with demand spikes for given acceptable wait profile and the probability of car sharing rises. Hence, each car spends less time idle (or travelling with just 1 passenger in peak periods) and more time earning revenue and defraying its fixed costs.

The simulation presents strong evidence that an on-demand, door-to-door, 24x7 public transport system based on an autonomous car fleet could be the best option for meeting Canberra's transport needs. The simulation demonstrates that a fleet of autonomous cars can provide a service that:

Summary

The costs of running ACTION

From the Territory and Municipal Services Annual Report, 2013-14:

• Each passenger boarding costs ACTION $7.96.
• As each MyWay journey requires 1.3 boardings on average (2012 data), each passenger journey costs ACTION $10.35.
• Fare revenues cover 15% of ACTION's operating costs.
• Hence, the rate-payer subsidy is $8.80 per passenger journey.
• The total cost of running ACTION for the year was $144.6M.
• ACTION's fare and charter income was $21.9M.
• The ACT Government paid ACTION $2.4M for special needs transport and $7.7M for concessional travel.
• The ACT Government paid ACTION an additional $92.3M "service payment" (also described as a Community Service Obligation in the budget papers.
• Even after these payments, plus fare and charter income, ACTION's "comprehensive deficit" was $28.6M

Hence, excluding the $2.4M for special needs transport in trying to create a "like for like" comparison, the rate-payers' contribution to ACTION was $7.7M (subsidy for concessional travel) + $92.3M ("service payment") + $28.6M (deficit funding) = $128.6M. To put this amount in context, the general rates revenue for 2013-14 was forecast as $338M, hence running ACTION (excluding special needs transport and fare income) costs rate-payers about 38% of their rates, or about $885 per household per year⁴.

The costs of catching a bus

Bus fares are cheap for the traveller: a weekday single adult fare using MyWay with a 5% discount (purchased by direct deposit) is just $2.70. The equivalent concession fare is just $1.34. Hence, the typical daily cost for a bus commuter is just $5.40.

The actual financial cost of the average journey is much higher: $10.35 per journey and $20.70 per day, but the difference is hidden in a rate-payer subsidy.

Other costs more difficult-to-measure include:

• Time costs of travellers. The door-to-door journey time for most passengers taking a bus is over twice that of motorists. In Hoang's scenario, this equates to 5 hours time less with his family each week.

• Inconvenience and inflexibility. As many of the scenarios describe, bus travel is neither convenient nor flexible. No viable bus system in Canberra could operate a 24x7 service, much less provide high frequency or on-demand and door-to-door services.

• Environmental: "tailpipe emissions". According to Estimating Greenhouse Emissions from Australian Capital Territory Travel Modes by Leon Arundell, 2012, the tailpipe emissions of a ACTION bus are about 92 grams of CO₂-e per passenger km:

  In 2009-10 ACTION buses used 7.227 million litres of diesel fuel and 2.554 million litres of compressed natural gas (CNG) to carry 16.94 million passengers (TAMS, 2010).
  Diesel fuel has an energy content of 38.6 GJ per kl and produces 69.9 kg CO₂-e per GJ.
  CNG has an energy content of 39.3 GJ/cubic metre and produces 57 kg of CO₂-e per GJ.
  On this basis ACTION's fuel use produces 25.22 million kg CO₂-e and carries 16.94 million passengers – an average of 1.49kg CO₂-e per passenger.
  ...
  Assuming an average passenger journey of 16.23 km, an ACTION bus emits 91.81 grams of CO₂-e per passenger kilometre.

  "CO₂-e" refers to "Carbon dioxide equivalent". Arundell's figure for a typical car is about 262 CO₂-e per passenger kilometre, almost 3 times more.

• Social and opportunity cost. What is the cost of social disadvantage caused to Jayne in not pursuing her CIT studies because transport makes it impractical? Or the reduced mobility of older citizens no-longer able to drive, or risks taken by people who shouldn't be driving?

The costs of running a car

The NRMA Car Operating Costs Calculator provides an estimate for the "whole of life" costs of running a car: depreciation, opportunity-interest (or financing costs), registration, CTP insurance, repairs, and fuel. It does not include costs of parking and no default cost of insurance is supplied.

Running the calculator on the top 5 cars sold in Australia in 2013 and providing a fixed estimate of $600 pa for insurance, shows the costs per km range from 53 cents for a Hyundai i30 (1.6l manual) to 71 cents for a Toyota Hilux (3l manual 4x2) to 100 cents for a Holden Commodore (3l manual sedan).

Assuming Hoang chose the most economical option (the Hyundai i30) and had an impeccable insurance record, his daily return commute from Palmerston to Parkes of 14.3 km would cost: 14.3km * 2 (return) * $0.53/km = $15.16. He would also need to pay for parking ($11.50/day purchased in advance), giving a total transport cost of $26.66 per day.

Other costs more difficult-to-measure include:

• Time spent cleaning, refuelling, arranging maintenance and driving.

• Environmental - greenhouse gases: around 260 CO₂-e per passenger kilometre as estimated by Arundall (and around 250 CO₂-e per km as estimated by the US EPA for an average US new car).

• Health and well-being. The Department of Infrastructure and Regional Development estimate the annual cost of road crashes as $1150 per person. Additionally, The Bureau of Transport and Regional Economics estimates that pollution from all motor vehicles (not just cars) causes between 900 and 4500 cases of morbidity and between 900 and 2000 early deaths each year.

• Traffic congestion. An analysis of the 2001 ABS Census data for Canberra by Gungahlin Drive Extension Review prepared by Scott Wilson Nairn Pty Ltd for the National Capital Authority in 2002 (page 16) estimated the car occupancy levels for journeys to work were 1.13. Such low occupancy rates contribute to traffic congestion. Morgan Stanley have estimated that autonomous vehicles could reduce productivity loss due to congestion by $US149 billion in the US annually, or about $US470 per person per year.

  The ACT's Minister for the Environment and Sustainable Development, Simon Corbell, stated in October 2014: "By 2020 congestion will cost the territory $200 million per year". The ABS estimate Canberra's population will be 437,000 (Series B estimate), giving an average annual cost of congestion of almost $460 per person.

• Land required for car parking. According to the ACT Government's Transport for Canberra - Transport for a sustainable city, 2012–2031 Table 5, the land value of an on-surface car-park is $92,829 in Civic and $17,707 in a town centre, with each car park costing $489 annually to maintain.

The costs of using an autonomous EV fleet (as simulated)

The average journey of 13.4km costs $3.80 in peak-period with the default simulation settings. Hoang's daily commute from Palmerston to Parkes is a little longer (14.3 km) and his return journey would cost $8.06. At an average per/km cost of 28 cents, the results of this simulation seem low in comparison with current car costs, but they are actually similar in $A terms to the September 2015 estimates produced by Deloitte of 31 US cents per mile (just under 20 US cents per km). The Deloitte modelling assumes lower capital costs based on light-weight, two-person vehicles, but much higher fuel costs (about 12 US cents per km).

[According to the ACT Government's Transport for Canberra - Transport for a sustainable city, 2012–2031, "40% of people travel less than 10 kilometres to work" (page 6), so for a significant number of commuters, their autonomous vehicle fare would be less than their current MyWay fare.]

For the "low usage" (ACTION equivalent) scenario, the autonomous system loses about $12M per year, or about 80 cents per journey (45K journeys per weekday, 32K journeys other days), bringing the true cost of Hoang's commute to $8.06 + $1.60 = $9.66 per day.

For the "high usage" (750K trips/day) scenario, the autonomous system makes about $78M per year, or about 30 cents per journey, bringing the true cost of Hoang's commute to $8.06 - $0.60 = $7.46 per day.

Currently, the ACT rate-payer directly and indirectly subsidises ACTION by about $128M per year (excluding special-needs transport). If that same subsidy were applied to the "low usage" scenario, then after the $12M loss running the autonomous car fleet, $116m per year would be available to subsidise the 15 million journeys undertaken. But total fare revenue in this scenario is only $46M, and the total cost of operating the system is only $58M. That is, all travel could be made free for a cost to the rate-payer of $58M, less than half of the cost to the rate-payer of supporting the current ACTION system. Passengers would have the added benefit of a door-to-door, 24x7 service.

However, under such conditions (free, excellent transport) demand could be expected to rapidly increase, possibly to the "high usage" scenario of 750K journeys per day, in which annual surplus from the system before subsidies is around $78M with total costs of around $691M and fare income of around $769M. A surplus of $78M could be used to provide almost 75,000 completely free off-peak journeys of average length (13.4km) every day to concession travellers, which is over 65% more than the total number of all journeys serviced by ACTION and the rate-payers would be saved $128M per year ($885 per household).

Unlike bus, the autonomous car is flexible and convenient, offering a 24x7 door-to-door service. Hoang can easily break his journey to pick up shopping on the way home; Mary can drop-off and collect her son, and call in at the supermarket.

The environmental costs, at least as far as "tailpipe emissions" are concerned, depend on the way the electricity they consume is generated. If brown coal is used, an EV may require more CO₂-equivalent per km than a 3l V6. If renewables are used, without considering the life-cycle costs of the renewables (energy required to construct and commission the renewable generator), tail-pipe emissions may be zero. It is not simple to generate a comparable equivalent CO₂-e cost for various forms of transport, but it is very likely that a transport system based on renewable energy will generate much lower carbon emissions than one based on fossil fuels.

The autonomous fleet simulation estimates car occupancy for two representative morning peak period journeys: from Kambah to Parkes and from Kambah to Civic East. A typical "high usage" run shows an average car occupancy of around 2.2 for both destinations, which is approximately double the estimated average current "journey to work" car occupancy of 1.13 people. By increasing car occupancy for commuter traffic, an autonomous fleet lowers congestion on major routes during peak hours.

A summary example of the direct financial benefits of an autonomous EV fleet

Consider Hoang, the hypothetical and typical commuter from Palmerston, deciding whether to support (with his wallet and his vote) an autonomous car fleet or support "business as usual" with his economical Hyundai and ACTION.

The financial advantages of the fleet are very significant: commuting in his car will cost him $26.66 each day; over 220 work days per year that's $5865. Further, his household as ratepayers contributes $885 to ACTION, giving a total cost to his household of $6750 (excluding the transport activities of other household members: maybe, like the majority of multi-person households, they have a second car). Using an autonomous car fleet will cost him $8.06 per day, or $1773. Assuming the "high usage" surplus is not returned to rate-payers but used to subsidise concession travel, Hoang's household is almost $5000 per year better off, not including the indirect benefits of reduced congestion and pollution, increased safety and improved productivity.

If Hoang's household has a second car, they may keep one car or share one with other families for long journeys during holidays, or they may use part of their $5000 savings to fly interstate or hire a car for occasional trips down the coast; with an autonomous fleet, it is easy and cheap getting his family to and from the hire-car depot.

A decision to replace a well-understood technology with one that is just coming of age is always difficult. Whilst no-one would want to be stuck in an age of candles and telegrams, few envy those who take responsibility for introducing change.

Nonetheless, because of the benefits it brings, "progress" is both inevitable and welcomed, and it is our responsibility to plan for its arrival and extract the greatest benefits we can for our community.

In planning for a transport system based on autonomous electric vehicles, amongst the major risks that need to be evaluated and issues requiring community discussion are:

Public transport in Canberra currently fails to meet the transport needs of the community.

By failing to provide a viable alternative to private cars it entrenches the many problems associated with private car travel including congestion, safety, pollution, land-use and cost. In failing to provide convenient transport to those unable to use or afford private cars, it entrenches social disadvantage which adversely affects the entire community. The current public transport service is also very expensive, requiring an annual subsidy from rate-payers of $885 per household, diverting much-needed funds from education, health and other community priorities.

It is important to understand that this problem has not been caused by ACTION and cannot be addressed with "tweaks" such as new timetables or raising parking fees. Rather, Canberra's low population and "bush capital" design forces a traditional public transport system to choose between a poor service with low-frequency services (long wait times especially out of peak periods) with sparse urban coverage (long walks to access and long routes and hence journey times), or unsupportable operating costs. The proposed light-rail does not help: indeed, by requiring the construction of a separate infrastructure it further dilutes resources available to be applied to the problem.

However, autonomous electric vehicles are now on the cusp of offering an alternative: a 24x7, on-demand, door-to-door service, providing mobility to all with the same convenience of the private car but at a fraction of the cost. This approach supports community objectives for reducing congestion, pollution and the land-use devoted to car-parks whilst improving transport safety.

At first glance, a fleet of autonomous vehicles sounds like science fiction, or at least 20 years into the future. However, the required technology is demonstrable and rapidly approaching commercialisation, in time frames that vary from two - five years (Audi, Tesla Motors, Google) to six - ten years (Nissan, GM, Ford). Independent assessments from KPMG, Morgan Stanley, Accenture, McKinsey, The Boston Consulting Group, Eno Center for Transportation, The Conference Board of Canada and others concur: it is very likely that autonomous vehicles will be successfully commercialised between 2017 and 2025.

The simulation implemented to model the financial viability of an autonomous vehicle fleet in Canberra provides strong evidence that under a wide range of car operational characteristics and journey demand levels, such a fleet can provide an extremely high level of service much cheaper than private cars and existing public transport alternatives. For example, a fleet of 23,000 autonomous sedans can service 750,000 "on-demand" door-to-door journeys per day, with over 98% of journeys starting within 1 minute of the request being received, and do so at fares much lower than undertaking the same journeys by private car, and whilst providing free transport for almost 75,000 journeys per day, all without requiring any rate-payer funding.

The implications for urban and transport planning are significant. Motivated by the myriad of problems caused by current transport options, the citizens of Canberra are currently considering a very large investment in a single-line light-rail system that will do almost nothing to address these problems, but at least has the virtue of "doing something".

It is hoped that a wider awareness of autonomous vehicles and the results of this simulation will prompt serious consideration of an alternative that meets all of the city's transport goals much more effectively, does so at much lower cost, and presents an opportunity for Canberra to take a lead in what will be one of the most transformative technological shifts of the 21st century.