Autonomous vehicle – Long, bumpy road ahead

As AV (autonomous vehicle) technology undergoes continuous improvement, the sight of AVs taking to city streets is gradually veering into the realm of probability. Major automobile manufacturers have begun to invest significant resources in AV research. It is estimated that the auto industry globally has invested $80 billion in AV development in the last three years alone, equivalent to the GDP of a mid-sized country.

No wonder that AV technology is quickly approaching maturity, with increasing investment in both hardware and software development. Today, we are on the verge of one of the most disruptive technological breakthroughs in generations. However, the success of AV adoption at scale depends as much on the ability of planners, designers, and policymakers at the Central, state, and city levels as getting the technology right. We can ill afford the consequences of overlooking any of the fundamental bottlenecks in the development of AVs.

Challenges in the way of vehicle autonomy

One of the major challenge lies in ensuring that consumers adopt autonomous vehicles for the long term.

Safety: For widespread autonomous vehicle adoption to become a reality, vehicle security is of paramount importance. Consumers want to be sure that as they’re sitting in the backseat working on their laptop, the vehicle will continue to coast down the highway safe and sound while their personal data remains private.

Part of the burden for autonomous vehicle developers is that there is virtually zero room for error. Millions of people die in road accidents each year in the world, but a single fatal crash involving a self-driven car on autopilot can all but eclipse them all. Nevertheless, the fact remains that if the automotive industry is going to make the adoption of AVs a reality, it will need to allay the safety fears of consumers.

Safety is the responsibility of automobile manufacturers, and they understand that there is a tremendous amount at stake. After investing billions of dollars, developers have already achieved a substantial amount of the technological breakthroughs needed to make AVs run, but what remains is finishing the crucial, final milestones left to assure rider and vehicle safety.

Auto manufacturers have to ensure visibility and the ability of AVs to autonomously analyse road situations in real time. And while developers work to overcome these technological hurdles, the issue of cybersecurity assumes prime importance. It is because we have to guard AVs against the possibility of someone maliciously changing the data running through the electronic control unit to manipulate it at his will. According to a research, the average AV is subject to more than 300,000 attempted attacks each month. And with more than 100 million lines of code in the average connected vehicle, there is no shortage of vulnerabilities for hackers to target.

For the “autonomous vehicle revolution” to become an everyday reality, cybersecurity infrastructure must be standard on all self-driving vehicles. Any holistic approach to autonomous vehicle safety must include embedded cybersecurity solutions, to keep the vehicle—and driver— away from the reach of those who can cause harm.

User Comfort: If AV is to become the future of transportation, consumers’ first experience of traveling in one must be pleasant, even enjoyable, otherwise it will be difficult to persuade them to repeat it. So, in focusing on the safety of AVs, the auto industry must not relegate consumer acceptance to secondary status.

This becomes essential because the process by which the AI (artificial intelligence) controlling an AV is trained relies on testing the algorithms involved for their ability to produce appropriate actions under all circumstances. If they are developed from the outset only for safety, then subsequently attempting to add additional requirements, such as a smooth ride or optimum path selection, could invalidate all the testing and require new algorithms to be developed from the scratch.

Travelling in an AV must not be equivalent to be left with a poor or inattentive driver prone to erratic behaviour such as failure to anticipate speed bumps across the road or potholes in the vehicle’s path, last-minute braking for stationary traffic ahead, incorrect lane positioning at complex junctions, and so on. Ultimately, self-driven vehicle must grow into an experienced and skilful driver who can anticipate and make allowance for both everyday hazards and more unexpected events, driving smoothly through even the most crowded and complex traffic scenarios, making no sudden changes of speed and direction. If we are to rope in consumers for AV on a regular or permanent basis, it is not enough to keep them safe—we must endow the AV with the capabilities of an expert driver. Configuring the AI to follow a path that minimizes speed and direction changes will improve passenger confidence, giving a reassuring impression that the vehicle “knows what it’s doing.”

The most valuable tool for developing these attributes in an AV is one of the new generation of highly dynamic driving simulators capable of realistically simulating sudden vehicle movements and responses to road surface changes.

The high level of immersion (realism) provided in such a simulator is necessary to ensure that a test driver reacts in a way that is fully representative of real life. The occupants’ subjective assessment of the AV’s behaviour, during a range of simulated driving scenarios, permits the algorithms to be thoroughly and repeatedly tested in a safe and controlled environment.

Modern simulators can even allow interaction between multiple vehicles, some of them under human control, and other road users, such as pedestrians. This enables the AV to experience the variability and unpredictability of human behaviour and trains the AI more rapidly than could be achieved through physical mileage accumulation on a track.

Advanced level data management

The importance of simulation in the development and deployment process enables robust testing and validation, and fleet orchestration to deliver a successful autonomous mobility service. On the vehicle sensor side, there is utmost need for rethinking and demystifying LiDAR as well as Smart Cities and V2X thermal imaging. Next in line are connected data initiatives, 5G connectivity, and the vital role of tech development ecosystems.

All data captured on the road are essential to developing a safe and fully automated driving experience, but there is an overwhelming amount of it to handle and process. Autonomous test vehicles typically generate between 5 and 20 TB of data per day, per vehicle. This includes cameras that tend to generate 20-60 MB/s, depending on the quality being captured from standard definition (SDs) to higher definition (HD/UHD), as well as sonar (10-100 kB/s), radar (10 kB/s), LiDAR systems (10-70 MB/s), and GPS (50 kB/s). The key is to ensure that sensors are collecting the right data, and the data are processed immediately, stored securely, made quickly available to analysis processes, and managed for long-term retention and recall for events such as system updates, regulatory changes, or litigation actions.

It is unthinkable to speak about AVs and data without mentioning how 5G will enable a new era of connected cars. Most AVs today are fitted with some form of in-vehicle communications system or Internet access. New developments in communications standards and networks have shown speeds of more than 10 GB per second and latencies of less than 10ms. All this leads to high speed, wide bandwidth, and extremely low latency.

Since 5G does not require a carrier, most of the AV data ingest and transfer can be done without wide area network coverage. Vehicle-to-vehicle and vehicle-to-infrastructure communications will move at a speed and capacity not possible with older generation networks. Thus, 5G expands the data potential tenfold. It promises to provide the speeds and data processing capabilities needed to mimic the timing of human reflexes. However, there is still a lot of work to be done to make that happen. As we get closer to approaching a 5G world, automotive manufacturers need to carefully handle AVs that will essentially exist as mobile data centres carrying and handling enormous volumes of data.

The move along the bumpy road to 5G and fully connected vehicles will be gradual. There are challenges that still need to be resolved, including the high frequency waves used by 5G. These waves do not travel far, get easily absorbed by surroundings, and can limit how quickly information travels between stations. We will see more base stations and more trials in the meantime, but a wide coverage footprint will take some time to achieve.

Manufacturers will have several choices to make when it comes to how they architect systems that receive and retain in-vehicle captured data in the future 5G world. More importantly, an infrastructure and workflow will need to be built that enables a smooth transition of data throughout its life cycle—from initial in-vehicle capture all the way through long-term retention of data that will be critical for future development.

For production connected vehicles and smart cities, 5G communications is expected to satisfy a significant portion of data-acquisition needs. However, the data do not need to be as rich as the video and video-like data associated with test vehicles.

For both test and production situations, all of the data need to end up somewhere and need to be available for a never-ending list of applications in areas such as intelligence, security, user experience, entertainment, commerce, and ongoing development. For these data, low-cost massive storage solutions are essential. These solutions must be highly scalable, they must include technologies for extreme data durability for long-term retention, and they must be able to make data easily searchable and accessible for those situations where it needs to enter back into high-performance processes.

As companies are setting up sound infrastructures for managing large datasets, they may have an occasional need to apply very high-performance processes to some of that data, but they do not have the CPU power required. In these situations, the practice of uploading a dataset to a public cloud for a limited time use, where resources can be provisions for short periods of time (cloud bursting), can prove to be the best overall method for providing cost-effective data management while meeting occasional demands for high-performance processing.

Integration in broader transportation ecosystem

While the technology is starting to take clear shape, AVs need to be seamlessly integrated into modern cities and lifestyles. They should not contribute to an urban dystopia of endless commutes through sprawling cities, but to a sustainable future where convenient and affordable mobility significantly contributes to the vitality of cities. It requires a collaborative approach between municipalities, urban planners, entrepreneurs, and engineers to harness the potential benefits of AVs.

In the long term, constructive public-private partnerships can lead to regulations that do more to support AVs than restrict them. We need integrated infrastructure that works with the vehicles rather than pose obstacles in the smoother transition to vehicle autonomy.

AV shuttle providers can bridge last mile connectivity to commuter rail stops to increase overall ridership and allow underperforming bus lines to be replaced by affordable and equitable AV services. As shared-ridership and -ownership models have been found to be a natural fit for AV technology, AV shuttles could be an ideal addition to existing transit systems.

Almost half of the land area in most cities today is devoted to roadways and surface parking lots. If shared AV shuttles become widely adopted, the space we now use for driveways could instead be used for backyards and larger houses, allowing considerable comfort along with urban convenience and efficiency. Car spaces can be transformed into people spaces, by reducing traffic, increasing safety, and encouraging more compact development. If door-to-door transportation were replaced with a network of carefully spaced designated pick-up and drop-off spots, we could see a more efficient overall transit network, activated neighbourhoods, and healthy local business communities benefitting from increased foot traffic. The reduction in traffic would mean that, rather than designing every street around cars, cities could include wider sidewalks, green alleys, and pedestrian streets.

Semiconductors shift to ensure greater auto reliability

Semiconductor content in vehicles is growing rapidly, both in volume and as a percentage of the raw parts cost of the vehicle, to serve safety, connectivity, and automation functions in the AV. Up from a few hundred, large-design-rule controllers, MEMS, power regulators, and other components a decade ago, a modern vehicle may now contain up to 8000 semiconductor chips that represent 20-30% or more of its cost.

As vehicles are tasked with automating more of the functions of driving, all semiconductor device segments are seeing growth, and AVs are now supposed to contain multiple complex SoCs (systems on chips), dozens of image sensors, and more memory (e.g., flash and DRAM) than a laptop or mobile phone

Increasing needs in advanced driver assistance systems (ADAS), AV technology and zero defect reliability are driving semiconductor development for automotive applications. As consumers and regulators demand more capability from automobiles, semiconductors have become a critical part of the advanced solutions. Many of the semiconductors are now part of ADAS that are critical to the function and safety of the vehicle, where failures cannot be accommodated.

The rapid push into autonomous driving capability accentuates the need for all chips to work together without incident to protect the safety of both the car’s occupants and others in the surrounding environment. This necessitates that zero defect systems be put in place across automotive semiconductor process control. Across the ecosystem, companies are recognizing the need to make fundamental changes in their historical approach to automotive chip manufacturing to respond to critical new trends and be successful in the automotive market.

The semiconductor industry has been discovering that automotive quality is quite different from consumer-grade quality. Reliability expectations for consumer-grade devices are orders of magnitude lower than the automotive market, allowing up to 10% failure rates within the first two years in a relatively tame operating environment.

Summing up

There are many hurdles to overcome before AVs are accepted and adopted across the globe, and the industry is right in prioritizing safety. However, by also accounting for passenger comforts in an AV at an early stage of its development will be critical to ensuring the overall success of the concept of vehicle autonomy.

While mapping, sensors, and intelligent algorithms will play a key role in realizing the safety benefits of AVs, the geometry of roadways, and the integration of intelligent infrastructure that facilitates communication between AVs, pedestrians, and cyclists are also the areas that merit attention of innovators and designers. Planning for roadways that can accommodate AVs can go a long way in realising the safety goals of AVs while facilitating a smoother transition to full automation. In fact, realizing the highest potential of AVs will require significant policy changes and coordinated investment in infrastructure. The innovators and entrepreneurs developing AV technology are at a critical moment to capture unprecedented change, but their ability to seize the moment depends on engaging in conversations with these groups to define shared goals and build partnerships from the outset of integration. It is these vital collaborations that can unlock the full potential of AVs to transform the future of mobility and lead to more vibrant, equitable, and sustainable cities.