5 Big Under-The-Hood Engineering Challenges In Building Autonomous Vehicles

Autonomous vehicle engineering is more than adding sensors to a vehicle.

popularity

Stories about autonomous vehicles are regular fare in the tech news cycle and usually include forecasts about the eventual ascendancy of self-driving cars. The Boston Consulting Group, for example, says that by 2035, 25% of all cars will have partial or full autonomy, with total global sales growing from near-zero levels in 2015 to $42 billion in 2025 and ~ $77 billion by 2035. In short few years since Google car, automakers new and old alike have announced their plans for commercializing autonomous vehicles starting as early as 2021. That’s disruption!

So what will it take to deliver on this future? Without a doubt, artificial intelligence with robust machine algorithms to comprehend road and traffic conditions and make appropriate decisions is the most critical. However, it will take a lot more than artificial intelligence to build commercially-viable autonomous vehicles. For starters, multiple space-constrained sensors (and the sensor fusion box) need to be integrated within vehicle electrical architecture and must function reliably in a harsh automotive environment. But autonomous vehicle engineering is more than adding sensors to a vehicle.

Here are five key areas which I think will need attention:

  • Vehicle sensor integration: Lidars, radar and cameras — three of the most common sensors for autonomous vehicles — are evolving quickly, with significant focus on reducing size and cost. These sensors will need to function reliably in all weather conditions. A big part of ensuring reliability in harsh environments is implementing thermal-conscious design of sensors’ signal processing electronics.
  • Zero-defect sensor fusion: Much like the human brain, the execution engine of an autonomous vehicle requires robust, real-time fusion of sensory data. Interpreting this data can be a life or death matter, so sensor fusion may well be the most critical aspect of vehicle autonomy. Essentially, a zero-defect fusion system is needed. Among the many questions when it comes to actually painting an error-free picture of the vehicle’s surrounding environment: Is it better to have distributed processing or completely centralized sensor data processing? Is there a power consumption target for fusion boxes and what is the best thermal management strategy? Design and implementation decisions hinge on such issues.
  • ECU consolidation: Vehicle autonomy will likely have increased demand for more infotainment, which in turn will drive more electronics/ECU consolidation. ECU consolidation may not always make engineering sense though desired. For example, demand for electronics cooling for such consolidated electronics may force significant size increases of components and systems. Such issues need to be taken into account from early design stage.
  • Electrical architecture and wire harness design: The wire harness is one of the heaviest parts of a vehicle. Increased electronics content and integration demand lead to tremendous design complexities. There can be more than 2 billion possible combinations for wires in a vehicle, a number that will only increase. Accordingly, one of the most critical engineering challenges is how to optimize wire harness design when connecting sensors to the sensor fusion box, and then integrating the sensor network with a vehicle’s overall electrical architecture.
  • Powertrain electrification: The rise of drive-by-wire systems and the proliferation of added onboard electronics require powertrain electrification. Building an autonomous system with an electric powertrain is not going to a simple additive engineering task, but rather will pose new challenges and dependencies. How will electric drive range be affected with the choice of an autonomous system? Are there new opportunities (and tradeoffs) for battery, motor, inverter sizing, especially since the behavior of the 90thpercentile driver need not be taken into account for autonomous vehicles (level 4 or 5)? There are many such powertrain engineering questions for autonomous vehicles.

Responses to these five challenges are readily apparent in the significant vertical integration and collaborations among automakers, sensor vendors and chipmakers. But is this enough? Especially when automakers want to roll out autonomous vehicles in the next 4 years and sensor/sensor fusion technology is still evolving?

I think not!

The entire evolving supply chain, from auto OEMs to technology vendors, will benefit from reliably evaluating design and operational implications for autonomous vehicles from the earliest design stages. Among the technology inputs most needed are frontloading simulation and characterization tools that can be easily used by vehicle engineers and designers, and not just by expert users. Benefits of such tools for ICE/hybrid/ EV powertrain design have been demonstrated. Now, extrapolating these benefits to autonomous vehicle engineering hints at significant reduction in time and cost for design and development.

Additional resources:
ADAS and the System Engineering Challenges
Engineering Edge