Singapore is counting on autonomous vehicles to power public transit. The vehicles are being tested on the road, under real-life conditions.
There is not much going on at the junction of Fusionopolis Way and Ayer Rajah Avenue: a few limousines, an electric bus, and a few pedestrians waiting at the light to switch. There is no indication whatsoever that ground-breaking work is being carried out here, in Singapore's One-North district. But then suddenly, a small electric car comes whizzing along with a camera system on its roof. The vehicle takes a right, weaves through the on-coming traffic (in Singapore they drive on the left) and must immediately swerve around an illegally parked motorcycle. But what's so special about it? The car does it all on its own, without the intervention of a human driver. The small white car is driving autonomously and is operated by the Future Urban Mobility initiative of the Singapore MIT Alliance for Research and Development (SMART), a partnership between the city state and the Massachusetts Institute of Technology (MIT) from the USA.
Real-life conditions accelerate progress.
Several projects are being carried out at One-North simultaneously, for example nuTonomy (another MIT offshoot) or A*star, a government research agency. The One-North district, founded specifically for start-ups and research companies has exactly the right amount of traffic for these tests. "We chose One-North because we wanted a test area with real traffic conditions," says Wee Shann Lam, Group Director for Technology and Industrial Development from the umbrella planning authority, the Land Transport Authority (LTA). At junctions without traffic lights, for example, the cars must learn how pedestrians move, says Lam. "There are no pedestrians on segregated testing grounds. Real-life conditions accelerate progress."
Singapore's greatest bottlenecks: there is a shortage of land and there are not enough drivers for buses and trucks
Rapid development is important to the government, because a solution for two urgent problems must be found: there is a shortage of land and personnel in the city state which has only 719 square kilometres and a population of 5.6 million. "12 percent of the state's surface area is already being used for road infrastructure. Apartments comprise only slighter more than 14 percent. We cannot continue to build more roads at the same rate," says Lam. Also there are too few local bus and truck drivers. For environmental reasons, an increase in individual traffic is not an option. "We believe that self-driving vehicles in local public traffic will help us considerably," Lam emphasises. "That is why shared and public autonomous vehicles are our starting point. We are not devoting our efforts to private autonomous cars." Singapore will continue to rely on the excellent MRT underground network for longer distances through the city for the masses. "And only the oldest line still has drivers," Lam smiles. Today, people are already recognising that the MRT is safe, whether a driver is controlling it or not.
The LTA is working closely together with the Ministry of Transport to promote self-driving vehicles. Together, they have developed regulations for testing autonomous vehicles. In the Committee on Autonomous Transport for Singapore (CARTS), for example, both agencies coordinate with economic authorities as well as with experts from the research and technology fields regarding regulations, their visions for the future, and industrial development.
With regard to the mobility of the future, Singapore has progressed further than Europe.
Jochen Siebert, Managing Director of the consultancy firm JSC Automotive Consulting, which specialises on the Chinese automotive market, also emphasises the city state’s leading role with regard to mobility. The German national recently moved from Shanghai to Singapore in order to devote himself better to the growing markets in South-East Asia. He has discovered that Singapore is approaching autonomous driving more deliberately and with fewer bureaucratic obstacles than other countries: with regard to the mobility of the future, Singapore has progressed further than Europe," also because the city state began to manage the flow of traffic at a very early stage.
About 20 years ago the government restricted the number of vehicle registrations. Furthermore, 15 years ago Electronic Road Pricing (ERP), an electronic toll system, was introduced. Payment is fully automatic; as soon as a vehicle passes through the system's white road gates. Vehicles don't have to stop; digital display panels show car drivers the current cost of using the road: the amount varies depending on the time of day.
Singapore is the ideal place to develop autonomous cars, explains Doug Parker, head of nuTonomy in the city state. "Singapore has great roads without potholes, no snow and no high hills. Traffic users keep to the rules and no one is on their smartphone all the time." And the government has a clear strategy.
This strategy includes autonomous taxis, buses, trucks and commercial vehicles. The city has progressed furthest with passenger traffic. For example, in the Gardens by the Bay park, the "Auto Rider" drives around for two hours every afternoon; the climatised electric vehicle can accommodate ten passengers and a ticket costs 3.20 euros.
The first autonomous minibuses are already on the road
Since spring 2017, an electric test shuttle with eleven seats and space for four people to stand has been travelling between Nanyang Technological University (NTU) and the CleanTech Park 1.5 kilometres away. It finds its way using GPS, four intelligent cameras which analyse traffic signs and traffic lights as well as eight lidar sensors whose laser beams sense their surroundings and detect obstacles. The vehicle's software evaluates the data immediately. The shuttle drives up to 40 kilometres per hour and according to Subodh Mhaisalkar, the head of the responsible research institute, it is being tested in "various scenarios such as stop-start traffic or on roundabouts".
Mobility on demand, not according to the timetable
Additionally, two test projects with 12 meter-long buses will be starting soon. "We wanted to have two different projects – so that, in the end, we can choose the best technology," says Wee Shann Lam, the development expert from the national planning authority. Just like anywhere else in the world, traffic in Singapore flows in waves: into the city in the morning and back to the suburbs in the evening. "We will always also need vehicles with a greater capacity for that." Self-driving vehicles that could be used for mobility-on-demand during the day would be ideal for this. That is why new tests with four autonomous mobility-on-demand minibuses will soon begin on the island of Sentosa, popular as a place of leisure, off Singapore's coast.
The objective with regard to the city state's freight traffic is, above all, to shift as much transport operations as possible to the night time and to scale down manpower needs. Having automated road-sweepers operate during the night will create more space on the roads during the day. In turn, autonomous trucks will save on drivers. An initial project is focussing on platooning: the first truck is steered by a person, whilst those behind merge into the convoy automatically. Here too, Singapore has approved two pilot projects by competing providers that are currently carrying out initial tests at their respective home locations.
In Singapore autonomous vehicles must be able to function completely independently: there should be as little infrastructure as possible
One special feature of Singapore is that the city state intends to function with as little infrastructure as possible. Developers must take that into account. "We intend for autonomous vehicles to know their own position and function with very little help," development expert Lam stresses. The government is relying on regulations and standards instead of infrastructure. Singapore wants to determine standards for tests and the evaluation thereof as soon as possible; for example in the field of coding, functional safety or networking. In 2016, a test centre was founded specifically to this end at the NTU technological university. At present there are no global standards. "So we are looking at local standards to begin with," explains Lam.
One task has already been partially completed: autonomous vehicles need digital maps with a degree of accuracy within two centimetres. Maps of this type are already available for One-North and it would take little more than a year to create some for the whole city.
We hope that in the future autonomous driving will free up city areas to make space for green areas, playgrounds and other uses.
The local newspaper, the Straits Times, sees the current projects as a precursor for a time when autonomous vehicles will be "as ubiquitous as smartphones are today." Wee Shann Lam, the development expert from the planning authority expresses Singapore's vision as follows: "We hope that in the future autonomous driving will free up city areas to make space for green areas, playgrounds, and other uses. People can move around or cycle there safely and at a leisurely pace. The vehicles will travel around one or two levels below – with stops, charging stations and workshops." It will be a while before this vision is realised. But Singapore has taken critical first steps on that path.
Mobility on demand is one of the key goals of future traffic concepts for Singapore. The idea is for commuters to be able to book autonomous shuttles or pods,via smartphone, i.e. to get home from the metro station. This type of "last mile" management aims to increase the attractiveness of public transport and thus decrease the need for people to use their own vehicles. What's more, Singapore promises to offer more mobility to older people, many of whom feel that the current system does not sufficiently accommodate their needs. Safety, efficiency and environmental friendliness are additional benefits which the new mobility service is expected to bring. What's more, with the autonomous fleet, less personnel is required – an important factor in Singapore, as the city state is already suffering from a lack of bus, truck and taxi drivers. Initially, an autonomous fleet of taxis is planned for use within one district, with the potential to increase the field of operation at a later time.
The startup, nuTonomy, is one of two companies that was asked to carry out tests to investigate the possibility of using autonomous short-distance taxis in Singapore. The company is currently testing six autonomous electric vehicles in the One-North district of the city. What is particularly interesting about these tests is the technology used in these vehicles. The nuTonomy founders, Karl Iagnemma and Emilio Frazzoli – both MIT experts for robotics and intelligent vehicle technology – have opted to not use machine learning whereby patterns are detected and compared with each other at various levels. Instead, formal logic is used. The NuCore software developed for this purpose is based on a hierarchy of rules, similar to the well-known "Three Laws of Robotics" by the science-fiction author Isaac Asimov. The first law states that robots must not harm humans, be it deliberately or through negligence. According to rule number two, all orders from a human being must be followed – unless these orders conflict with the first law, that is. Meanwhile, the third rule demands that robots protect themselves – unless this law conflicts with rule one or two. Creating such a system is complex. "You first have to select the right rules, and these must then be transcribed into computer code," says Doug Parker, head of the Singapore location.
At nuTonomy, the most important rule for the vehicles is that they must not come into contact with soft objects, i.e. humans or animals. Secondly, they must not touch any hard objects. Subordinate rules include, for example, the requirement to maintain trajectory or speed. And as a general rule: rules based on laws and ethics take precedence. Less important are what Parker calls "cultural rules". What is the minimum distance to another vehicle? Should the vehicle let pedestrians cross the road in front of it, even if there is no sign post to this effect? That's because the expectations of humans on the behaviour of a car driver – and thus also of an autonomous vehicle – differ in every society. The software uses a planning algorithm called RRT*, a variant of the so-called "rapidly exploring random trees" (RRT) principle, whereby a space is established with randomly selected, branching paths – a system which is also used for autonomous robot movements. Each path in RRT* corresponds to a specific solution path, providing an answer to the sensor data imported from the vehicles. The software evaluates these paths and selects the route which best suits the prescribed rule hierarchy. The advantage of this is that the system also works in highly complex situations. "And the chaos in urban areas is intense," Parker explains.