Mapping could hold to key to mass adoption of ADAS

The development of ADAS technology brings with it a host of new challenges, from simulation data and real-world feedback, to the need for collaborative partnerships and common formats. While there are possible solutions arising from a vastly increased number of tests and map data, there are also questions over how the industry could tackle the growth issues that have arisen. 

Key trends

From ADAS to total autonomy, driver assist technology is moving forward very quickly, says Kirk Steudle, director at Michigan Department of Transport. “It’s widespread both geographically and across companies (technology companies, automobile companies, etc). There is a recognition that there is going to be an enormous need to develop, test and validate this technology, off-road and on-road.”

Exactly how to validate these systems is an ongoing topic of discussion. In Europe, the focus is on forward collision warning and automatic emergency braking as these are due to become part of the NCAP safety evaluation. There is also concern about lane keeping assist as more data reveals that customers are choosing to turn the function off owing to the system being too sensitive, says Sam Abuelsamid, senior research analyst at Navigant Research. “Unfortunately there doesn’t seem to be any consensus on how best to evaluate any of these systems,” he says.

It remains to be seen how the National Highway Traffic Safety Administration (NHTSA) will develop standards for autonomous vehicles in the US. “Under the current administration, it’s unlikely that any new regulations of any kind will move forward,” says Abuelsamid. “Given that no one wants regulations that hamper innovation as the technology develops, any rules or standards need to be technology agnostic and focus on validating a base level of functionality.”

In both the EU and the UK there seems to be recognition of the long-term value of automated driving and laws are being changed to permit on-road testing, says David Alexander, senior analyst at Navigant. How to manage the constant evolution of the technology is a key trend and discussion point. “There is recognition that the need to develop the systems will continue even after the first introduction of the technology; as new versions of automated components, systems and vehicles are developed, they will need to be tested and validated in off-road and on-road scenarios,” says Steudle.

Alexander agrees. “Validation in a controlled environment can only go so far, and vehicles need to be evaluated by a wide range of drivers under real road, traffic, and weather conditions.”

Possible solutions

There are many challenges arising from a vastly increased number of tests and map data streams for ordinary roads. All our interviewees agreed that a solution could be huge amounts of simulation, necessary to validate system and sub-system level functionality, says Abuelsamid. “It will be nearly impossible to provide adequate validation from traditional road testing, although that type of testing will still be needed to evaluate durability, interoperability with other automation and human driven vehicles and different environmental conditions,” he says.

Traditional automaker development processes involve exhaustive internal testing before the product is released to customers. For highly automated vehicles, carmakers will probably decide to operate their own fleets or only supply vehicles to fleet operators who commit to collaborate with them in ongoing evaluation and development, says Alexander. “Systems and vehicles must be designed for quick and easy modification to be able to adapt to new developments.”

In regards to maps, collaborative data is a possible solution and is already starting to happen with partnerships across multiple data suppliers and carmakers. This will accelerate in the coming years and keep costs of the data collection as manageable as possible, says Abuelsamid.

“Accessibility of test and map data will continue to challenge, and be of interest,” says Steudle. “There’s no doubt that there will be significant value to the data and maps generated by advanced systems. We are keen on seeing if there is an agreeable business model out there where the data and maps can be shared with public transportation agencies.”

Growth issues

The main growth issues for ADAS include the lack of map data for simulation and developing how to teach AI with real-world data or simulation data. How do we solve these? Standardising one way of working would be the ideal thing, say our experts. “Consortiums to collect, aggregate and share data would be the ideal thing. However, for this to be viable, we need common formats and means of normalising data,” says Abuelsamid. “Different companies have different sensors, mounted in different ways and adjustments must be factored into any shared data set to provide an environment model that can be used by everyone.

For now, each company has been creating HD maps of the locales where they are planning to test and then using that in their control system. This has in part limited where tests take place. For production deployments shared map models will likely be needed. Numerous companies are working on how to aggregate data from vehicles and distribute this including Mobileye with its Road Experience Management, TomTom’s RoadDNA and CivilMaps, among others. Since everyone is operating in the same physical environment, an open source model with contributions from everyone and made available to everyone would probably be ideal but that is unlikely.

There is another possibility, says Alexander. “The alternative is continued growth of AI so that vehicles learn to drive as people do, by analysing what is in front of and around them and making decisions based on real-time sensors and being less dependent on mm accuracy from digital maps.”


Leave a comment

Your email address will not be published. Required fields are marked *