driverless Archives - Artificial Intelligence

System trains driverless cars in simulation before they hit the road

aiuniverse — Tue, 24 Mar 2020 07:58:46 +0000

Source: news.mit.edu

A simulation system invented at MIT to train driverless cars creates a photorealistic world with infinite steering possibilities, helping the cars learn to navigate a host of worse-case scenarios before cruising down real streets.

Control systems, or “controllers,” for autonomous vehicles largely rely on real-world datasets of driving trajectories from human drivers. From these data, they learn how to emulate safe steering controls in a variety of situations. But real-world data from hazardous “edge cases,” such as nearly crashing or being forced off the road or into other lanes, are — fortunately — rare.

Some computer programs, called “simulation engines,” aim to imitate these situations by rendering detailed virtual roads to help train the controllers to recover. But the learned control from simulation has never been shown to transfer to reality on a full-scale vehicle.

The MIT researchers tackle the problem with their photorealistic simulator, called Virtual Image Synthesis and Transformation for Autonomy (VISTA). It uses only a small dataset, captured by humans driving on a road, to synthesize a practically infinite number of new viewpoints from trajectories that the vehicle could take in the real world. The controller is rewarded for the distance it travels without crashing, so it must learn by itself how to reach a destination safely. In doing so, the vehicle learns to safely navigate any situation it encounters, including regaining control after swerving between lanes or recovering from near-crashes.

In tests, a controller trained within the VISTA simulator safely was able to be safely deployed onto a full-scale driverless car and to navigate through previously unseen streets. In positioning the car at off-road orientations that mimicked various near-crash situations, the controller was also able to successfully recover the car back into a safe driving trajectory within a few seconds. A paper describing the system has been published in IEEE Robotics and Automation Letters and will be presented at the upcoming ICRA conference in May.

“It’s tough to collect data in these edge cases that humans don’t experience on the road,” says first author Alexander Amini, a PhD student in the Computer Science and Artificial Intelligence Laboratory (CSAIL). “In our simulation, however, control systems can experience those situations, learn for themselves to recover from them, and remain robust when deployed onto vehicles in the real world.”

The work was done in collaboration with the Toyota Research Institute. Joining Amini on the paper are Igor Gilitschenski, a postdoc in CSAIL; Jacob Phillips, Julia Moseyko, and Rohan Banerjee, all undergraduates in CSAIL and the Department of Electrical Engineering and Computer Science; Sertac Karaman, an associate professor of aeronautics and astronautics; and Daniela Rus, director of CSAIL and the Andrew and Erna Viterbi Professor of Electrical Engineering and Computer Science.

Data-driven simulation

Historically, building simulation engines for training and testing autonomous vehicles has been largely a manual task. Companies and universities often employ teams of artists and engineers to sketch virtual environments, with accurate road markings, lanes, and even detailed leaves on trees. Some engines may also incorporate the physics of a car’s interaction with its environment, based on complex mathematical models.

But since there are so many different things to consider in complex real-world environments, it’s practically impossible to incorporate everything into the simulator. For that reason, there’s usually a mismatch between what controllers learn in simulation and how they operate in the real world.

Instead, the MIT researchers created what they call a “data-driven” simulation engine that synthesizes, from real data, new trajectories consistent with road appearance, as well as the distance and motion of all objects in the scene.

They first collect video data from a human driving down a few roads and feed that into the engine. For each frame, the engine projects every pixel into a type of 3D point cloud. Then, they place a virtual vehicle inside that world. When the vehicle makes a steering command, the engine synthesizes a new trajectory through the point cloud, based on the steering curve and the vehicle’s orientation and velocity.

Then, the engine uses that new trajectory to render a photorealistic scene. To do so, it uses a convolutional neural network — commonly used for image-processing tasks — to estimate a depth map, which contains information relating to the distance of objects from the controller’s viewpoint. It then combines the depth map with a technique that estimates the camera’s orientation within a 3D scene. That all helps pinpoint the vehicle’s location and relative distance from everything within the virtual simulator.

Based on that information, it reorients the original pixels to recreate a 3D representation of the world from the vehicle’s new viewpoint. It also tracks the motion of the pixels to capture the movement of the cars and people, and other moving objects, in the scene. “This is equivalent to providing the vehicle with an infinite number of possible trajectories,” Rus says. “Because when we collect physical data, we get data from the specific trajectory the car will follow. But we can modify that trajectory to cover all possible ways of and environments of driving. That’s really powerful.”

Reinforcement learning from scratch

Traditionally, researchers have been training autonomous vehicles by either following human defined rules of driving or by trying to imitate human drivers. But the researchers make their controller learn entirely from scratch under an “end-to-end” framework, meaning it takes as input only raw sensor data — such as visual observations of the road — and, from that data, predicts steering commands at outputs.

“We basically say, ‘Here’s an environment. You can do whatever you want. Just don’t crash into vehicles, and stay inside the lanes,’” Amini says.

This requires “reinforcement learning” (RL), a trial-and-error machine-learning technique that provides feedback signals whenever the car makes an error. In the researchers’ simulation engine, the controller begins by knowing nothing about how to drive, what a lane marker is, or even other vehicles look like, so it starts executing random steering angles. It gets a feedback signal only when it crashes. At that point, it gets teleported to a new simulated location and has to execute a better set of steering angles to avoid crashing again. Over 10 to 15 hours of training, it uses these sparse feedback signals to learn to travel greater and greater distances without crashing.

After successfully driving 10,000 kilometers in simulation, the authors apply that learned controller onto their full-scale autonomous vehicle in the real world. The researchers say this is the first time a controller trained using end-to-end reinforcement learning in simulation has successful been deployed onto a full-scale autonomous car. “That was surprising to us. Not only has the controller never been on a real car before, but it’s also never even seen the roads before and has no prior knowledge on how humans drive,” Amini says.

Forcing the controller to run through all types of driving scenarios enabled it to regain control from disorienting positions — such as being half off the road or into another lane — and steer back into the correct lane within several seconds. “And other state-of-the-art controllers all tragically failed at that, because they never saw any data like this in training,” Amini says.

Next, the researchers hope to simulate all types of road conditions from a single driving trajectory, such as night and day, and sunny and rainy weather. They also hope to simulate more complex interactions with other vehicles on the road. “What if other cars start moving and jump in front of the vehicle?” Rus says. “Those are complex, real-world interactions we want to start testing.”

The post System trains driverless cars in simulation before they hit the road appeared first on Artificial Intelligence.

Driverless truck startup Starsky Robotics folds: CEO shares tough autonomy home truths

aiuniverse — Sat, 21 Mar 2020 07:13:09 +0000

Source: en.brinkwire.com

Driverless truck startup Starsky Robotics is shutting down, but not before sharing some cold hard truths about the autonomous driving industry. Founded in 2015, Starsky proposed a combination of self-driving and remote control for a fleet of next-generation trucks, saving full autonomy for the highway.

Rather than build a driverless truck that could handle every situation, Starsky’s plan was to mix autonomous systems with human teleoperation. On the highway, a relatively controlled environment, the truck would drive itself. That way, the demands for skilled operators would be significantly reduced.

In trickier situations, however – effectively “the first and last mile,” as Starsky explained it – a human driver would take over the controls. They wouldn’t be physically present in the truck, mind. Instead they’d use remote controls to pilot the rig from a distance.

Back in 2019, Starsky demonstrated the first fully-unmanned truck to drive on a live, public highway. Now, though, the company is shutting down. In a blunt post-mortem of what went wrong, founder and CEO Stefan Seltz-Axmacher blamed results-hungry investors, unexpected difficulties with getting the AI right, and the fact that safety just isn’t sexy for Starsky’s problems – and the problems that he predicts will impact the self-driving industry as a whole.

“There are too many problems with the AV industry to detail here,” Seltz-Axmacher writes, “the professorial pace at which most teams work, the lack of tangible deployment milestones, the open secret that there isn’t a robotaxi business model, etc. The biggest, however, is that supervised machine learning doesn’t live up to the hype. It isn’t actual artificial intelligence akin to C-3PO, it’s a sophisticated pattern-matching tool.”

The issue, he explains, is that matching – and eventually exceeding – human drivers’ abilities with edge cases is much tougher than most realized. Everyday driving in reasonable conditions is fairly low-hanging fruit; that can be achieved relatively rapidly. Developing a system that is capable of reacting safely to unexpected situations, however, is far trickier, and as you refine the self-driving AI you also set yourself the challenge of finding increasingly specific risk models with which to test.

Adding to that problem is the fact that, while safety is often cited as a primary concern for people when asked about whether they’d get into an autonomous vehicle, it’s actually a tough thing to get people excited about. The same, Seltz-Axmacher says, goes for investors. Starsky spent almost two years working on safety engineering, but “the problem is that all of that work is invisible,” he writes.

“Investors expect founders to lie to them,” the Starsky founder explains, “so how are they to believe that the unmanned run we did actually only had a 1 in a million chance of fatality accident? If they don’t know how hard it is to do unmanned, how do they know someone else can’t do it next week?”

At the end of 2019, the company’s attempts to raise more money fell flat. It’s currently seeking to sell off its patents as the company breaks apart. Seltz-Axmacher says he sees real autonomy still being 10 years out; “no one should be betting a business on safe AI decision makers,” he concludes.

The post Driverless truck startup Starsky Robotics folds: CEO shares tough autonomy home truths appeared first on Artificial Intelligence.

The Digital Single Market: A focus on robotics and artificial intelligence

aiuniverse — Fri, 18 Oct 2019 07:58:02 +0000

Source: openaccessgovernment.org

The Digital Single strategy of the European Commission sets out to “open up digital opportunities for people and business and enhance Europe’s position as a world leader in the digital economy.”(1) This article will briefly examine a part of that by looking at the work of the Robotics and Artificial Intelligence (Unit A.1).

We know that the Unit sets out to push forward the development of a competitive industry in robotics and artificial intelligence (AI) throughout Europe. Certainly, this includes industrial and service robots plus the growing field of autonomous systems from drones and driverless vehicles to computing and cognitive vision.

Also, the Unit encourages the wide uptake and best use of robotics and AI in all societal and industrial fields. The Unit takes responsibility for the European Commission’s participation in the contractual PPP on Robotics and for the implementation and development of the relevant strategic industrial agenda.

As well as managing RD&I projects within Horizon 2020, the UNIT keeps up-to-date with legal and ethical issues in the field of robots and autonomous systems, for example, aspects related to the impact of automation and robotics on jobs and work environment, as well as liability and safety.

Currently, Juha Heikkilä is Head of Unit at Robotics and Artificial Intelligence (Unit A.1) (2). Juha joined the European Commission in 1998 and today works in their (3) Directorate-General for Communications Networks, Content and Technology (DG CONNECT). This is the department “responsible to develop a digital single market to generate smart, sustainable and inclusive growth in Europe.” (4)

In July 2019, we find out that the European Commission launched a call to develop a vibrant European Network of AI Excellence Centres. Under the Horizon 2020 Work Programme 2018-2020, proposals for this can be submitted up to 13th November 2019. Europe has important potential to lead technological advancements in AI, with a strong research infrastructure and world-class community of scientists at their disposal. It is, therefore, crucial that the crème de la crème research teams in Europe collaborate to combat significant technological and scientific challenges.

The Commission is looking ahead at a long-term effort to unify the European AI community and make the region an AI powerhouse. To achieve this, they believe that two actions are needed:

Research and Innovation Action to mobilise the best researchers into networks of excellence centres to reach a critical mass on important AI topics.
Coordination and Support Action to enable exchange between the selected projects, as well as other relevant initiatives.

These aforementioned actions should create synergies with the industrial sector and, “foster an ecosystem of R&D resources, expertise and infrastructure (in areas such as HPC, robotics equipment, IoT infrastructure).” (5)

In other news, a significant piece of robotics news from the European Commission came in late 2018 when they awarded €66,000,000 to robotics projects that will help digitise companies throughout the European Union (EU). As part of the Digitising European Industry Call of Horizon 2020, the EU’s research and innovation programme, one coordination support action and four projects have been awarded. We read more about this on the European Commission’s website.

“They will all help small and medium-sized enterprises (SMEs) adopt new technologies in the robotics and artificial intelligence area. Nearly half of the money dedicated to these Digital Innovation Hubs (DIHs) projects will be dispatched to local companies by involving them in mini-projects or experiments.”

The four awarded projects are:

DIH^2: A network of 26 DIHs to reach no less than 170 DIHs.
DIH-HERO: A broad pan-European network of DIHs specialising in healthcare robotics will be established.
TRINITY: The aim here is to create a network of multidisciplinary DIHs consisting of research centres, university groups companies that cover a wide array of topics.
RIMA: This sets out to establish a network of 13 DIHs on robotics to facilitate the uptake of maintenance and inspection and maintenance technologies. (6)

The topics briefly discussed are just a few examples of the excellent work of the Robotics and Artificial Intelligence (Unit A.1) within the European Commission and we look forward to future developments in this most exciting area of work. It’s a really important aspect of the Digital Single Market but it’s not the only part of it.

The post The Digital Single Market: A focus on robotics and artificial intelligence appeared first on Artificial Intelligence.

Argo.ai and Carnegie Mellon to found driverless vehicle research center

aiuniverse — Tue, 25 Jun 2019 06:36:27 +0000

Source:- venturebeat.com

Argo.ai, a Pittsburgh, Pennsylvania-based driverless car startup founded by former executives from Google’s and Uber’s autonomous technology divisions, today announced that it’s teaming up with Carnegie Mellon University to form a new center for autonomous vehicle research: the aptly named Carnegie Mellon University Argo AI Center for Autonomous Vehicle Research.

Argo.ai says it’ll pledge $15 million over five years to fund faculty leaders and support graduate students conducting studies in pursuit of their doctorates. Additionally, the company says it’ll provide Carnegie Mellon students engaged in autonomous vehicle research access to data, infrastructure, and platforms like Argoverse, a curated corpus of more than 300,000 vehicle trajectories and 290 kilometers of recorded road lanes.

In a blog post, Argo.ai principal scientist and associate professor at Carnegie Mellon Deva Ramanan said that the Center will investigate smart sensor fusion, 3D scene understanding, urban scene simulation, map-based perception, imitation and reinforcement learning, behavioral prediction, and software validation as they relate to driverless vehicle technology. More broadly, it’ll pursue projects to help self-driving cars overcome hurdles like such as winter weather or construction zones, and Ramanan expects its work will “spur engagements” on topics like safety policy and ethics.

“While the team at [Argo.ai] sees a pathway to achieve initial commercialization opportunities for self-driving technology, there are still advancements required to be able to perceive and navigate autonomously in the most complex, open conditions with dramatically lower compute power,” wrote Ramanan, who added that all of the Center’s findings will be reported in open scientific literature. “And until we’re able to do so at scale, the visionary benefits that have been spelled out for society won’t be achieved.”

Ramanan will serve as the Center’s faculty leader along with Simon Lucey, an associate professor at Carnegie Mellon University’s Robotics Institute specializing in computer vision. The team’s other founding members include John Dolan, David Held, and Jeff Schneider.

“We are thrilled to deepen our partnership with Argo.ai to shape the future of self-driving technologies,” said Carnegie Mellon president Farnam Jahanian. “This investment allows our researchers to continue to lead at the nexus of technology and society, and to solve society’s most pressing problems. Together, Argo.ai and [Carnegie Mellon] will accelerate critical research in autonomous vehicles while building on the momentum of [Carnegie Mellon’s] culture of innovation.”

The Center follows on the heels of Argo.ai’s existing collaboration with Carnegie Mellon and Georgia Tech, but it’s worth noting it’s not the first of its kind. Intel last October announced that it would launch an Institute for Automated Mobility in Arizona, which will combine three state universities; the Departments of Transportation, Public Safety, and Commerce; and companies working on automated cars, trucks, and drones.

Argo has a close relationship with Ford, which in February 2017 said it would invest $1 billion in the startup over the next five years to help it achieve its goal of producing a self-driving vehicle fleet by 2021. This made Ford the company’s largest shareholder and enabled Argo to hire 200 additional employees, many of whom were Ford employees working in the R&D department on a virtual driver system.

Autonomous hardware and software stacks remain Argo’s core projects, along with the high-definition road maps and virtual driver system that will eventually slot into Ford’s self-driving vehicles. Ford has previously said it intends to launch a self-driving taxi and delivery service by 2021.

A.T. Kearney: Get used to competing in digital disorder era

DEAN TAKAHASHI@DEANTAKJUNE 24, 2019 09:01 PM

But the global management consulting firm predicted in a report that by 2030, a new digital era will emerge. The trajectory of the global regulatory environment for technology as well as the extent to which the Internet remains open or balkanized will determine the contours of this period.

To be positioned for the future digital era, businesses must engage in a strategic digital transformation. A.T. Kearney’s Score framework presents a road map for this process.

A.T. Kearney said there are different possible futures, with fears growing about a new “digital cold war” and the “splinternet,” where the Internet becomes more balkanized. This is forcing companies around the world to shift strategies on everything from procurement to customer engagement.

In a new report by A.T. Kearney’s Global Business Policy Council, Competing in an Age of Digital Disorder, the firm said that companies can no longer be passive observers of the digital revolution.

Instead, they must actively adapt to the present disorder while also preparing for the future digital order by embarking on strategic end-to-end digital transformations.

Much attention is focused on the “techlash” nature of new policies on key issues such as consumer privacy, data protection, and anti-competitive practices. But many governments are now seeking to strike a balance in policies that both maximize digital’s upsides and mitigate its downsides as they prepare to regulate the digital environment for the first time. Whether those
governments are able to deftly strike such a balance will influence companies’ ability to use digital technologies effectively in the coming years.

“This cycle of innovation, adoption, and then regulation is consistent with previous waves of technological change,” says Paul Laudicina, chairman of A.T. Kearney’s Global Business Policy Council and co-author of the report, in a statement. “Today, the intense regulatory debate regarding digital technologies is creating a high degree of uncertainty about how the policy environment will evolve.”

After providing a richly researched background on the opportunities and pressure points facing societies, governments, and businesses in this period of digital disorder, the study then offers four scenarios for the digital order that will emerge.

The scenarios are based on two political uncertainties that are unfolding:

Regulatory activity. The extent to which governments in key markets around the world impose new regulations on technology companies and the use of digital technologies more broadly
Digital environment. The extent to which the digital economy is a globalized whole, characterized by extensive cross-border digital flows, or an islandized environment, fragmented into different country-level or regional blocs

“These scenarios are designed to be compelling and plausible visions of the future that challenge and test executives’ capacity to anticipate and plan for their companies’ digital strategies in the coming years,” said Erik Peterson, managing director of the Global Business Policy Council and co-author of the study, in a statement. “In fact, some aspects of these scenarios, such as the
emergence of a digital ‘cold war’ between major global powers and early indications of a ‘splinternet,’ are already playing out in various markets around the world.”

Finally, the study argues that companies cannot be simply spectators of the ongoing digital revolution. Instead, executives will need to guide their organizations through strategic digital transformations across a variety of business functions.

“Companies must adapt to the emerging digital order across strategy, customer experience, operations, risk management and compliance, and employees and culture—our Score framework,” said Courtney Rickert McCaffrey, manager of thought leadership for the Global Business Policy Council and co-author of the study, in a statement. “To compete in the 21st-century digital economy, companies must embark on end-to-end digital transformation in all SCORE

The post Argo.ai and Carnegie Mellon to found driverless vehicle research center appeared first on Artificial Intelligence.