Autonomous Trucks Hitting the Road Using Camera-Based Technology
The retail industry is growing faster than it has in 15 years, accelerated as of late by the pandemic and consumer willingness to order everything online for delivery — from groceries and alcohol to toothbrushes and batteries. This massive growth of shipments has put great pressure on the trucking industry, which reports a shortage of 900,000 qualified truck drivers in the US right now, and 2–8 hour delays at ports to pick up shipments, with no relief in sight. So it shouldn’t come as a complete surprise that autonomous trucks just might be hitting the roads sooner than autonomous passenger vehicles.
The reason for this is that deploying autonomous trucks is a much simpler task when limited to point-to-point routes on highways. This simplification is reasonable because the majority of truck miles occur on highways where driving complexity is minimized. On a highway, a truck needs to be able to stay in its lane, be aware of other vehicles around it, and be able to handle unforeseen situations like work zones, fallen obstacles, and traffic situations like stopped vehicles.
TuSimple, a company that develops autonomous trucks, has partnered with trucking company Navistar and UPS to conduct autonomous vehicle testing on the road in both Arizona and Texas. Currently, these tests require a human driver to be present; however, their plan is to soon eliminate that and conduct live road testing without a person in the truck. Some of their recent tests that included transporting watermelons from Arizona to Oklahoma City shaved enormous delivery time off their route — what typically took 24 hours was done in 14 hours, proving the advantages that autonomous trucking can provide to the industry.
Companies are receiving strategic investments from investors looking to capitalize on the fast growing autonomous trucking market. Kodiak Robotics, a company that develops self-driving truck technologies, recently received $125M to add trucks to its fleet and further develop their autonomous service capabilities. Investors are predicting that autonomous trucks will surpass the market value of robotaxis and are expected to reach $2,013.34M by 2027. Investors see autonomous trucks as a means to create efficiencies, fill the gap when there are labor shortages, and enhance safety.
In fact, the savings alone for deploying L4 trucks is enough to justify the influx of investment dollars, not to mention the increase in operating hours (autonomous trucks can operate 20 hours/day). Today, a truck costs about $1.69/mile to run, 43% of which is the cost of the driver. Removing the requirement to have a human driver saves considerable money, improves the productivity of each truck, and importantly, alleviates the industry’s dependence on qualified truck drivers.
“The move to autonomous trucks has the potential to transform an old, stodgy, Industrial Age industry into a high-tech one,” said Greg Aimi, Research Director at Gartner. “Tomorrow’s truck drivers being hired as technology-enabled supply chain professionals could reverse the shortage problem and might also justify higher pay for a more desired job. What is certain is that autonomous trucks will become reality at some point in the future. It’s a matter of when, not if.”
However, some of the challenges facing the autonomous trucking industry are the difficulties in sensing small objects at a long range, which has been an obstacle for the industry in deploying L4 trucks. Additionally, trucks can have high shock and high vibration issues (because they have such large engines) which may pose a problem for some autonomous sensors.
Today, 3D vision technology that uses multiple cameras, otherwise known as stereo vision, are independently mounted on the truck to provide long range, high resolution, real-time 3D sensing that works better than LiDAR, especially in the critical “small object at long range” scenario. Based on standard automotive-grade cameras that today ship numerous units/per year, these vision systems produce dense 3D point clouds with highly accurate distance estimates to each pixel, making the measurements that the camera senses much more accurate. This data provides the detail needed to navigate complex situations and avoid potentially lethal obstacles.
Long-range sensing capabilities are crucial for trucks, as they take 350 meters to stop in a smooth manner and 160 meters for an emergency stop when traveling at 65 mph. 3D vision technology can accurately sense objects up to a 1000 meters away, providing more safety and sensing than a human driver, or LiDAR could achieve.
As the demand for autonomous trucks increases, a direct impact from retail and economic shifts, implementing the right sensors to decrease delivery times, improve efficiency, and safety also becomes critical. Long-range camera-based technology just might get L4 autonomous trucks on the roads faster than the industry once predicted.