Self-Healing Software eliminates human-error and human-apathy

In the last year, every conversation about automotive software - every speech, every blog and every presentation - all consistently highlight the growing amount of code in the car. With this increased amount of code comes more software bugs, more recalls and the risk of more hacks - both from black hat and white hat hackers.

This conversation is consistent across countries and regions. Where differences lie, however, are in how regulatory organizations hold companies and people responsible to fix the problem that caused the need for a recall.

The details on the different levels of responsibility will be reviewed in a moment. Before we get there, it is important to note that it is the contention of Aurora Labs that the responsibility should be taken away from the human being and the guarantee that the software gets to a safe state should be offloaded to Self-Healing Software and in-vehicle software management solutions.

Now, back to the regional regulatory story. U.S. law requires auto manufacturers to alert owners when a vehicle has a recall, so that owners can take their car to a dealership, where the defect will be fixed for free. However, there are no laws requiring the owners to actually follow through and have the defect repaired.

Under US federal law, the cost of recalls for cars 15 years old or less is covered by automakers. But despite the free repair of potentially dangerous problems, there are an estimated 46 million cars with unfixed recalls currently on the road. According to the National Highway Traffic Safety Administration (NHTSA) and the Government Accountability Office, only 65 to 70 percent of vehicles subject to a recall are repaired within the 18-month period during which automakers provide recall completion data.

In the UK an estimated one in 13 cars is subject to an outstanding recall. This is a concern that has both the Driver and Vehicle Standards Agency and the Ministry of Transportation investigating ways to remedy the problem. One solution under consideration is taken from the German model which issues owners repeated warnings if their car has an outstanding recall. If the recall has not been fixed when tested for their biannual road worthiness test (or General Inspection as it is called in Germany), the car fails the test and cannot be driven on the road. The challenge here is to connect the car ownership database with the car maintenance systems to create a unified database of the vehicles VIN, their current HW and SW state, and their owners.

Enter OTA Updates. Forecasts predict that the growing amount of software in the car will reach 40% of the total BOM by 2025, and while hardware related recalls will require a visit to the mechanic, software recalls are remotely updateable (over-the-air, or OTA). While this method is far more efficient than a visit to the mechanic, here too there is no guarantee of a 100% completion rate. Current OTA update solutions transfer responsibility of the recall software update to the drivers by informing them of the update and requesting that they perform certain actions before initiating the update. Many drivers are either intimidated by the often technical nature of the messages and the actions required while others ignore the warnings under the miss-assumption that if it doesn't seem to be broken, it doesn't need to be fixed.

Aurora Labs' In-Vehicle Software Management solution approaches remote OTA updates with a novel approach. Using machine learning algorithms, the update file is generated in such a way that it can be installed with zero need for driver intervention and with zero downtime. This is the same user experience we have come to expect from consumer software such as our web browser and smartphone apps. The Chrome browser updates every 4-6 weeks without requiring user intervention and apps update daily. In the case of a vehicle recall, should it be deemed a mandatory recall, the vehicle will be able to self-update without requiring the driver to be part of the decision process, guaranteeing the vehicle continuously runs the latest, safest and most secure software.

As the industry transitions to autonomous cars and mobility services, the need to guarantee that the vehicle is always up-to-date will only increase. It is only natural that as Artificial Intelligence is used to enable self-driving cars, that it also enables self-healing vehicles.

Creating self-healing cars with human-like intelligence

Editor's Note: Anyone born after 1970 should take the time to click on the links pretty funny stuff.

Oscar Wilde said life imitates art. In our industry, technology imitates the imaginations of TV producers from the 60s, 70s and 80s. Maxwell Smart of Get Smart was the first to use a mobile phone - so mobile it was embedded in his shoe. The Jetsons were the first to have a "smart house" and KITT from Knight Rider was the first intelligent car based on the "Knight Industries Two Thousand" (KITT - acronym and all) processor.

 

Here we are heading into 2019 and expectations are that the number of mobile phones around the globe will pass the 5 billion mark; robots are being used in homes and enterprises, and based on the automotive industry's transition from hardware to software-centric offerings - we are on the cusp of the intelligent car.

AI and Machine Learning will be cornerstone technologies used to develop an intelligent and independent car, allowing autonomous vehicles to conduct and continuously learn how to do human-like activities. The car will soon be driving itself allowing people to do other things during the time they are being transported. Even imagine a pedestrian walking down the street and the autonomous car knows enough to flash the headlights - or blink his/her eyes - to let the person know it is safe to walk across the street.

Also like the human body, imagine this car with software so sophisticated that the software itself detects anomalies, stops bad processes from causing the car to malfunction and seamlessly updates with new features and functionality - fixing itself like the human body heals itself.

Similar to the realization on Get Smart's mobile phone and the Jetson's smart home, Aurora Labs believes that advances in AI and Machine Learning weaved into our Self-Healing Software solution will help make the vision of the intelligent self-healing car a reality.

Updating, and still waiting

Yay! There is a new software version available for my car head unit. I wasn't informed about it by the OEM or my dealer, I found out about it on one of the car owner forums I am on. I followed the link and downloaded the 2.16GB(!) image to my pc, transferred it to a USB drive and went out to my car. 30 minutes in, 34% and this is going to take a long time! It wouldn't be so bad if I could leave it running in the car but, and I quote from section 18 of 32 from the 20-page official update guide: "Press and release the clutch pedal (Manual Transaxle) or the brake pedal (Automatic Transaxle) and proceed with the following steps without stopping. NOTE: If the ignition is left in the ACC position it will automatically switch off after 25 minutes. Performing this step resets the timer to ensure the vehicle does not switch off while installing the reinstallation package."

Not exactly the user experience I have come to expect from any of the many connected consumer devices I possess. This process wouldn't be any different if I had taken my car to the garage - they too would have had to follow this same 32 step procedure. Software management including steady updates of features and functionality need to become the norm and not the exception. Our cars need to start behaving the same way as our smartphones - seamlessly updating, without requiring user intervention, to give us new features and improved functionality on an ongoing basis. I'm working with a team that are developing a clientless OTA update technology that will enable all and any ECUs in the car to be updated instantly without requiring user intervention and with zero downtime. This technology will enable the car software to seamlessly and constantly improve, making me feel more positive about buying another car from the same brand next time around.

Now if you'll excuse me, I've still got 12 more manual steps to do before I complete this update and I need to keep my foot on the brake!

Stepping up to meet the challenge of software complexity

The 2018 Automotive Elektronik Kongress in Ludwigsburg was not about autonomous cars or any other technology that comes with a 10-year engineering horizon. When the industry's most senior people that are actually tasked with building the cars and technologies of the future met in one room, they talked about stepping up to meet the challenge of software complexity and continuous integration & development.

Over half of the 25 companies who presented at the 2018 Automotive Elektronik Kongress talked about the critical need for Software Lifecycle Management and Continuous OTA Updates. These included: Elektrobit, Bosch, TTTech, BMW, Escrypt, Audi, Deutsche Telekom, Volkswagen, Harman, Here, Continental, Intel, and others.

There was wide agreement in the Kongress that Software Lifecycle Management is essential to speed up a time to markets of new product functionality, fix software bugs, enable new mobility and post-sale business models, and continuously maintain the car's security.

In interesting discussions I had with many of the attendees, there was a widespread agreement of the need for a next-generation OTA Update technology that enables the updates to happen with zero downtime, without requiring additional memory and a higher BOM. Improved user experience is paramount for OTA updates to become widely used throughout the vehicle.

OEMs have already embarked on the journey to not only introduce new technology but also change the organization and processes to meet this need for a continuous software release world.

Perhaps best summarizing this change in Volkswagen, Dr Rolf Zoeller quoted Charles Darwin: "It is not the strongest or the most intelligent who will survive but those who can best manage change."