Interview: How can advances in AI and sensor fusion make low-cost ADAS a reality?

My keynote will showcase new research that will help technologists accelerate their roadmaps for autonomous technology and achieve improved reliability at a significantly lower cost.

The current economic crisis has emphasised the vital need for more affordable ways to develop ADAS technologies. Without this, it will be impossible to hit the system cost points that make it possible to have autopilot features in mid-range vehicles.

My talk will focus on emerging approaches and solutions for this very challenge. Participants will gain insights on new approaches to deep learning and the fusion of low-cost sensors, which can dramatically reduce the cost and time to market of self-driving technologies, whilst achieving radical performance improvements.

The keynote will also highlight advances in AV simulation and emerging technologies to mitigate the effects of sensor distortion in poor weather, whilst upscaling depth information from low cost sensors through synthetic data and training environments.

As a consultancy focused on technology development, we help companies across the automotive industry achieve the most complex breakthroughs at rapid speed – channelling deep expertise in sensing (optics, radar, optical systems), wireless connectivity (cellular, satellite to short range), AI and edge computing.

Over the years, we have developed automotive technology for many vehicle OEMs, Tier 1 and component suppliers, including radar, telematics, connectivity devices and audio systems. This heritage includes creating Cambridge Silicon Radio, which became market leader for automotive grade Bluetooth and Wi-Fi technologies.

More recently, we are helping players across automotive, off-highway, Mobility as a Service (MaaS) and smart infrastructure sectors develop technologies required for self-driving systems. In many cases, this involves creating more innovative ways to lower the cost and improve deep learning performance within ADAS systems.

We’ve just developed a highly innovative AI model that fuses outputs from low-cost depth sensors and high-resolution camera input. This enables vehicle technology developers to create high-resolution depth data at vastly reduced cost than available today.

My keynote will be the first time this is seen beyond our labs and we’re going to enable each participant on the webinar the opportunity to log into a web-based demonstration to try the technology for themselves after the presentation.

Once in the demonstration environment, each participant will be able to change the sensor fusion inputs and see for themselves how deep learning can dramatically improve the available depth information for a particular RBG image. It should be a lot of fun and be a truly immersive experience for everyone to try.

AI and inference accelerators are now maturing as a technology solution for automotive systems. This will enable vehicle OEMs and suppliers to benefit from the cost savings that will surely come as the technology matures.

With more automotive grade semiconductor devices offering edge inference support, I believe the wider automotive supply chain will see a growing opportunity to generate new sources of revenue from the use of AI in devices embedded in sensors and in-vehicle engine control units. Acceptance and adoption of these technologies will also grow as standards for AV technology matures.

I see significant opportunities for the automotive market as a result of these advances. However, to capture the full value we must continue to invest in engineering innovation to surface new sources of value.

Join Thomas Carmody’s session at ScaleUp 360° Advanced Driver Assistance Systems Europe:

Session title: Revolutionising AI and sensor fusion for autonomous vehicles
Follow this link to sign up now for free: Online registration
Signing up automatically gives you access to all webinar sessions at ScaleUp 360° Advanced Driver Assistance Systems Europe, June 30 - July 01.