Remco Timmer, SVP Product Management, Head of Automotive Solutions at HERE Technologies, discusses issues around trust and safety – as well as the role for mapping – in developing driverless vehicles.
What is your sense of where we are on the journey to self-driving autonomous vehicles?
With any new technology, the first reaction of the public is often mistrust or even fear, and the road to building trust can be long. Self-driving technology is still early on in that journey, and the stakes are high as assisted and autonomous driving are critical safety functions with human lives at stake if the technology fails. As it stands, the general public does not feel comfortable with the idea: the World Risk Report found just 27% would be comfortable in a self-driving vehicle. Research in the Journal of Risk Research found that trust, rather than knowledge will be key to driving acceptance of self-driving technology.
How important is ADAS tech in this?
Car makers must build trust through the gradual adoption of Advanced Driver Assistance Systems (ADAS). Our collective knowledge of the road network, efficiently encoded into maps, combined with robust, scalable training and validation in simulations, and transparent communication to facilitate collaboration between humans and machines all play vital roles in this transition.
By introducing simpler, low-level automation systems as defined by the Society of Automotive Engineers (SAE), car makers can accustom users to the many advantages of self-driving systems. These benefits include smoother and more relaxed rides, as well as improved safety and efficiency, which will nurture confidence amongst the public. . Consumers need to become familiar with how vehicles can seamlessly and continuously support their driving.
Looking ahead then, what do you see as the next critical steps to achieving progress?
On the transparency front, I believe there are three critical steps.
Being able to see what cars ‘understand’ about the world around them will be fundamental to building confidence as car manufacturers gradually introduce autonomy, starting with systems such as Intelligent Speed Assistance. Over time, this progresses towards SAE level 3 Autonomy, where cars can handle many driving tasks, allowing drivers to take their hands off the wheel and even their eyes off the road at certain times, as defined in the operational design domain (ODD).
The ODD definition is encoded onto a map layer. It outlines where, under what conditions an ADAS feature is or is not available. This is the first key element to transparency. Drivers need to understand in advance what level of ADAS support is available. The navigation system can use it to plan the most supported route, and the vehicle can communicate in advance to the driver when they can gain more or when they have to regain control of the vehicle.
The second critical step is to have drivers educated on how the car works, which will be reassuring. It is similar to sitting next to someone you believe to be an inexperienced driver; you need reassurance that they know what they are doing. For that to happen, it helps for the driver to see and understand what the car ‘sees’ and processes about the environment around it. By demonstrating its complete understanding, drivers can relax and trust the vehicle. The car’s perception stack, the sensors the vehicle ‘sees’ through, such as camera, radar and lidar, work in harmony with high quality map data as the vehicle needs to understand the environment beyond what it can ‘see’. The perception stack’s field of view might be obstructed (by a corner or elevation) or impaired by other vehicles or weather, however the vehicle also needs to be aware of non-observable information that is encoded in the map, like road rules and regulations.
The third step is for users to understand and feel comfortable with what the car is about to do next. Not only what it can do next, as outlined in step one, but more specifically its manoeuvres. The car must constantly communicate its intent so that users know, for example, when it is about to overtake or change lanes. Clarity in communication from the car is essential to avoid everything from drivers feeling surprised or disoriented which could lead to anxiety and even motion sickness.
Transparency is key but we need to ensure that the level of detail can vary so that the driver is not needlessly bombarded with information, offering more detail in heavy traffic and less detail while cruising on the highway. Offering more to inexperienced users and less to experienced users.
Where do maps fit in?
Maps will be vital to delivering the awareness that the vehicle and the drivers need to fully trust autonomous vehicles. Sensors underpin the ADAS systems that help drivers and keep them safe, but alone, these sensors are not enough: the map is what sees ‘beyond’ the sensor range, delivering contextual awareness. With live maps delivered over the air, it becomes easier to communicate to the user what the car ‘understands’ about the environment. This helps build trust by reassuring the driver that they are safe, but also that the technology is working to make sense of the world around them.
Maps are far more than just a navigation tool. They can help a vehicle ‘understand’ the world around it. By working as a ‘dynamic canvas of collective knowledge of the road network’ maps can be used by a vehicle to both depict (provide context) and understand the world. As a consequence, drivers enjoy smoother manoeuvres and more efficient rides. The dynamic live map is a canvas for collaboration. For instance, another car driving further down the road might see accidents or other hazards, and transmit that data to the dynamic map, so the next driver can be aware of the danger.
Testing and validation for driverless systems is a challenge though isn’t it?
Cars should already be prepared for unforeseen circumstances, before they ever hit the public roads. Behind the scenes during the ADAS development stages, the ADAS systems are trained and validated in simulation environments. This will help to improve the performance of self-driving and driver assistance systems, reassuring human drivers that their cars will perform well in complex and unfamiliar situations.
By tracking where human drivers prefer to take control or tend to take over the controls or feel uncomfortable, car makers can identify situations where the systems might need additional training. These events can be map matched so that the OEM can easily identify hundreds of similar scenes and scenarios to use in their training. Training and validation in simulated environments can also help recreate challenging situations, allowing ADAS and autonomous systems to train, respond and improve over time. Doing this virtually enables the OEMs to do this at a massive scale well beyond what would have been possible with real world drives. The map is used as a source to ensure that the vehicle is trained in realistic virtual environments.
But it is not just machines that provide feedback to boost the performance of self-driving systems: it’s also vital to incorporate human feedback. Usage analytics allow car makers to home in on the situations in which human drivers feel uncomfortable, and where drivers need more support. By monitoring this information, manufacturers can ensure that if one user does not enjoy the experience, the next user is better served, redefining the vehicle’s operational design domain (ODD) which encodes what level of support drivers get.
When it comes to trust then, there are risks that can be mitigated as this technology is further developed?
Trust must be earned, and car makers have a long road ahead and risk damaging it quickly if they don’t adopt the right technology. Mapping software has a vital role to play to add to resilience and safety, as do systems that clearly communicate what is happening, and a gradual approach is vital to building this trust over time.
Once drivers understand the work that maps and broader software is doing to keep them safe, they will be more inclined to trust their increasingly self-sufficient vehicles, and ready to enjoy the roads of tomorrow.
Remco Timmer, VP Product Management and Head of Automotive Solutions at HERE Technologies
Remco Timmer is an experienced product and technology leader with a track record of driving innovation across industries, including automotive, consumer electronics, and healthcare information technology. With expertise in product management, innovation strategy, and both hardware and software development, he has successfully led multidisciplinary teams, integrating user experience (UX) design with strategic product development. Currently serving as Vice President of Product Management – Head of Automotive Solutions at HERE Technologies, Remco leads product strategy and technology initiatives that shape the future of location intelligence in the Software-Defined Vehicle era.
