Embark
San Mateo, Calif.-based Embark might be the most impressive self-driving truck company on the list. The startup, which unveiled its technology in February 2017, recently completed five-day, 2,400-mile drive from LA to Jacksonville, Fla. There was a human safety driver inside the truck for the entire trip, but reports said “the vast majority of the driving was autonomous,” with “hours at a time with no disengagements.” When there was a disengagement, they usually lated “only a few seconds.”
Embark is different from other self-driving trucks in that its trucks don’t rely on high-resolution, detailed maps of its route. Embark’s system use a combination of radars, cameras and LiDARs to collect data, which is then processed by deep neural networks (DNNs) to allow the truck to learn how to drive. According to Embark, this approach reduces the cost and development time needed to establish new routes.
Embark is primarily using its self-driving trucks for long stretches of highway driving. It relies on human truckers to drive the freight between warehouses and highways, where Embark’s computer takes over.
What would be accepted today by both drivers and trucking companies is driver assistance systems. Hardware to help with safe following, and especially, systems to help keep the speed safe as road conditions change, while alerting the drivers to conditions ahead. Reducing fatigue while improving driver awareness would be accepted by both sides of the issue. Who could argue with improving both safety and profits?
Here is another question, which is, how many different responses does the truck control computer have for any given situation, either actual or possible? The lack of the correct response is why the UBER car killed that woman in Florida. The car had only two choices and both were wrong! When the computer vision system noticed the woman next to the road it should have moved away, either into the next lane, or at least to the far edge of it’s own lane. That is what a smart human driver would do. . But it is certain that the computer only saw a clear lane and thus it was not in a position to avoid hitting the poor woman. If the car had been on the left edge of it’s lane it would have either missed her or just grazed her, but instead it saw no potential problem and so went on with it’s program of staying in the lane center. This is the intrinsic flaw in computer driven cars and there is no way to fix it. The best approach would be to stop wasting money before hundreds of folks die!