Autonomous Vehicles (AVs) are leading the new-age revolution of the automotive industry. Equipped with disruptive technologies, automakers across the globe are aggressively investing in building their AV or self-driving car portfolio. Despite offering a promise of reduced road fatalities, cheaper car rides and an easy mobility option for individuals with visual impairment or with physical disability, autonomous vehicles are still largely limited to special trial programs.
According to research, by 2025 we’ll see approximately eight million autonomous or semi-autonomous vehicles on the roads across the globe. But before merging into roadways, self-driving cars will first have to progress through six varying levels of driver assistance technology advancements. Defined by The Society of Automotive Engineers (SAE), the different levels of driving automation range from 0 (fully manual) to 5 (fully autonomous). While humans are responsible for monitoring the driving environment in the initial three levels, automated system monitors do the job in the latter ones.
Vehicles belonging to level 5 do not require human attention and are not even expected to have steering wheels or acceleration/braking pedals. They will be free from geofencing, able to go anywhere and do anything that an experienced human driver can do. While fully autonomous cars are undergoing testing in several pockets of the world, none are yet available to the general public. As of now, commercial use of fully autonomous vehicles have largely been restricted to repetitive missions in limited or monitored areas like that of shuttles for airports.
A close examination of the barriers in the AV-rollout throws light on the sluggish expansion of the industry which is currently growing at a rate of 16% per year1. While most auto players are equipped with hardware that can offer the requisite computational power, a critical roadblock in the full-fledged development of AVs is the absence of essential software to support the hardware. A major problem area that needs resolution, with respect to software, is ‘object analysis’- a feature that can enable a vehicle to detect an object and recognize what it represents2. The primary challenge in ‘object analysis’ is detection, which can be influenced by sudden physical movements, geographical or weather conditions. Additionally, it would be difficult to achieve a sensor fusion to validate the type and existence of an object because of the types of data that need to be compared by the camera, radar and the Lidar system.
Another crucial problem area which needs to be addressed is that of ‘decision-making systems’. It is imperative for AVs to make spontaneous decisions for a myriad of scenarios and undergo intensive training to mimic human decision-making skills. Categorizing different scenarios and building a database of ‘if-then’ rules for every
potential scenario is not feasible with the existing technology. Furthermore, AVs must be equipped with a fail-safe mechanism which lets the car stop without putting the passengers or the people around at risk.
However, there is no standardized procedure to check every software state and build safeguards for every potential outcome. That’s why to test the combination of the onboarded system for detection, it is necessary to test the vehicle for really long distances and in various real as well as simulated conditions. According to Waymo, a wholly owned subsidiary of Alphabet Inc., their autonomous vehicles have driven 20 million miles on public roads and more than 10 billion miles in computer simulation since 2009.
Legal liability is one more constraint that is hindering the expansion of AVs: What is the consequence of an accident procured by the car if the driver is not responsible anymore? Is the car maker or the software vendor legally liable? How to protect these vehicles from cyberattack? What can happen if the vehicle cannot be geo-localized for even a split second? Is the surrounding environment ready to take control of the vehicles through intelligent/cloud infrastructure? In most of the cases, these are still open questions.
It will be immensely challenging for a single player to address the software-related challenges and develop solutions in the short-term and accelerate the growth of the industry. Hence, an integral step towards developing software which can utilize the full potential of autonomous-vehicle hardware is for automakers to form more industry partnerships apart from the ones that are already existing as the boost on digital technology arrived suddenly, and is far from a traditional OEM’s DNA. The integrated development framework built through partnerships with technology start-ups and OEMs is significantly boosting the development of AV technology. As examples Daimler and Baidu are cooperating in Autonomous Driving and Vehicle Connectivity; BMW is working with start-up DOVU to pilot Blockchainfor tracking mileage in leased vehicle; Alibaba has partnered with Daimler, Audi and Volvo for AI voice assistance; Daimler, Bosch and Nvidia have teamed up to bring robot-taxis to silicon valley.
Most of the challenges mentioned earlier, can be addressed through intensive R&D coupled with long-test and validation periods. For instance, through rigorous testing and validation, developers can address the fail-safe mechanism, supplement the database of if-then rules with an AI engine and make inferences for scenarios which are not covered by the database.
-
Vehicles belonging to level 5 do not require human attention and are not even expected to have steering wheels or acceleration/braking pedals. They will be free from geofencing, able to go anywhere and do anything that an experienced human driver can do
Sanjeev Varma, Chairman - Altran India
Globally, the US is leading the AV revolution with heavy lobbying from auto giants. It is one of the few countries which has a federal guidance for Automated Driving Systems, put forward by the U.S. Department of Transportation and the National Highway Traffic Safety Administration (NHTSA). India, on the other hand, has maintained distance from the AV ecosystem. Recently, the Minister for Road Transport and Highways had announced that self-driving cars will not be rolled out in India as it might create unemployment among drivers across the country. Some research firms are also of the opinion that it might take another generation to make the rollout of AVs viable for low automated regions such as India.
Several Indian start-ups are currently in the process of developing AV products for trucks, minibuses and cars, with a focus on exporting to other countries in some cases. Although AVs plying on Indian roads may seem like a distant dream at the moment, we are fast progressing towards finding unique and niche applications for AVs where they can help deliver additional value, safety and efficiency, such as in-plant logistics, in-campus movement and public transport. With our country’s strength lying in innovation and technology, it is not completely unrealistic to imagine India someday becoming the leading supplier of AV technology to the world, thus incubating an entire Silicon valley for Autonomous Vehicles.
