Attention

Attention

This website is best viewed in portrait mode.

Publication Name: Automotiveworld.com
Date: June 15, 2023

AVs are a certainty - the path towards them is not

AVs are a certainty - the path towards them is not
While Tata Elxsi’s R&D works to bring solutions to major AV challenges, some answers come only in the form of waiting or major gambles. By Stewart Burnett

Although autonomous vehicles (AVs) are inching closer to becoming a reality, there are several technological hurdles that must first be overcome. Tata Elxsi, part of Tata Group, is at the forefront of autonomous research and development (R&D) in South Asia, working diligently to bring solutions that address problems at both a regional and global level. Having been a player in the industry from an early stage, it is keenly aware of the challenges the industry faces, and what kind of solutions it must deliver—building them via the company’s dedicated advanced driver-assistance system (ADAS) digital platform.

Some of the biggest challenges the company has identified include how to effectively distribute the vast computational workload, how to utilize 5G technology, and how to overcome technological disparities between different players in terms of both hardware and software. While the company has made strides towards overcoming some of these challenges, others depend on factors entirely outside of its control, thus drawing out the timeframe. However, this also gives it more time to develop and build out new solutions.

Achieving universality

Ashwin Ramachandra, Tata Elxsi Digital Service Practice Head, emphasizes to Automotive World that “AVs are an ongoing project—we are still in the early days. ”Some of the technologies necessary for their success remain in the conceptual stage, or fall far short of the scale required. One of the biggest issues, he remarks, is the disparity of parts different AVs will utilize, and how to create a universal solution for all of them.

To this end, the company has developed a middleware solution called Autonomai, which connects the base hardware components of the vehicle to higher functionalities such as cloud computing and connectivity—for instance, vehicle-to-vehicle (V2V) or vehicle-to-everything (V2X) communication. “We start at the silicon level and build upwards towards middleware,” says Ramachandra. Through Autonomai, he believes companies will have the means to develop autonomous applications that are interoperable and compatible with any kind of in-vehicle hardware.

AVs promise to revolutionize the transport sector forever, but there are significant technological hurdles which must first be overcome.

Middleware is, however, only part of the solution—one that can only lay the foundations for resolving the challenge of AV computational power. Ramachandra believes that “the trick is in sampling and averaging.” Having developed are presentative picture from data samples, whether from sensors or the battery, he explains that a determination can be made as to what needs to be sent to the cloud, and what can be handled in-vehicle. However, he concedes that this depends on 5G being universally available—including in rural areas—as 4G lacks sufficient bandwidth to push out the necessary volumes of data.

Creating solutions in advance

The slow roll-out of 5G has not stopped Tata Elxsi from experimenting with its use – cases. Ramachandra highlights an experiment the company recently conducted with a 5G testbed to determine whether driver monitoring could be off-loaded into the cloud. “Surprisingly, it worked,” he remarks. By sending the data feed from the driver-facing interior camera to the edge of the network, the company was able to show that monitoring driver alertness can be handled with vehicle hardware. Although he emphasizes it is only a proof-of-concept if put into practice this would make the computational workload easier to manage. Unfortunately, it may not happen any time soon—McKinsey has estimated that by 2030, high-band 5G will only be available to roughly 25% of the global population.

Ramachandra remains an optimist about timelines. In the case of V2X, he claims to be “fairly bullish” that it will become a reality in the next three to five years and that, when it does, the company will be ready. One example he cites is creating alerts for incoming emergency vehicles: “The way things are at present, you have to rely on your hearing to know an ambulance is nearby.” Instead, he posits, V2Xcommunication could alert vehicles several miles in advance of the ambulance’s passing, giving them time to create a passage, and thereby improving the efficacy of emergency services as a whole.

Acknowledging uncertainty

Although these examples are promising, many fundamental questions about what technology to adopt remain unresolved. Different solutions for AV connectivity are emerging on a regional basis, with a cellular 5G variant seeing increasing uptake in China and North America, and a non-Wi-Fi direct short-range communication alternative already becoming prevalent in European and Japanese R&D. However, these technological variances may turn out to be just a short-term complication on a journey towards eventual consolidation. “It’ll sort itself out,” states Ramachandra,“ A path towards unification is there”.

AVs are an ongoing project—we are still in the early days

Instead, effectively distributing the computational workload generated by AVs is likely to remain the biggest challenge. Lucid Motors has previously estimated that the combined data from sensors could reach as high as 40 gigabytes per second. Processing this will depend on either particularly robust in-vehicle hardware or a heavy dependence on the cloud—neither of which, Ramachandra observes, are feasible.

The company does, however, see a solution in the emergence of neural graphics processing units and artificial Intelligence Markup Language (AIML). Both of the technologies promise to enhance processing ability while reducing the overall amount of code required—the latter is even capable of using natural language interfaces, a development that could, for instance, allow passengers to issue commands to their vehicle via direct speech. Ramachandra believes that up to 40%of all in-vehicle computation could be handled by AIML alone—something he acknowledges is a “big gamble” for the company, which is strategizing under the assumption this will be the case. Although nothing is certain, he still concludes that such an approach could be tantamount to a “holy grail for this industry.”