2 answers
Voc
Student
Landrum, South Carolina
42
Questions
Updated
986 views
How can you make a car drive with no buddy in it?
I like stories with cars are sentient machines #mechanical-engineering
Login to comment
2 answers
Joseph Neilson
Radiation Detection Physicist
245
Answers
Oxford, England, United Kingdom
Updated
Joseph’s Answer
There was a question just like this on here a couple of months back, but I can't seem to find the link now. I think they were coming at it more from a filmmaking and special effects perspective though, so whilst literal self-driving was mentioned, the answers were more about filming tricks like greenscreens, a hidden driver, or towing the car on winches.
For your question, we're going much more in the direction of current "self-driving" cars and autonomous vehicles. We don't quite have the sort of sentient machines like you might have seen in Pixar's Cars, Herbie, Knight Rider or Thomas the Tank Engine, but there are a lot of companies working on self-driving technology. There's already a lot of mainstream vehicles with advanced lane-keeping features and similar that are doing part of the driving autonomously, Tesla's "self-driving" features are pretty advanced, and we've had some autonomous trains and airport shuttlepods for a while now.
If you're asking how that sort of thing works - it's all about sensors like cameras and LIDAR to "see" what's going on around the vehicle and software to respond to what it's measuring in the appropriate way. A lot of regular human-driven cars are now "drive-by-wire" - controls like the throttle and brake aren't directly connected to a physical throttle or brake pad any more, they're sensors and switches that feed your input into a computer, and the computer sends out control actions to motors and solenoids that make the actual moving parts of the car go. For self driving cars, you simply add in another computer sending the inputs digitally based on sensors and cameras rather than a human pressing buttons and switches.
As to a sentient machine like the stories, that sort of thing is a long way off. We don't have any artificial general intelligence that's anywhere close to sentience. There's some pretty cool AI research like GPT-3 and neural networks, but it's still got a long way to go. Even if we were close, an AI clever enough to be sentient would be capable of a lot more than just being a driver - it starts bringing up questions of whether it would want to just be driving around, and what sort of rights an AI might have.
For your question, we're going much more in the direction of current "self-driving" cars and autonomous vehicles. We don't quite have the sort of sentient machines like you might have seen in Pixar's Cars, Herbie, Knight Rider or Thomas the Tank Engine, but there are a lot of companies working on self-driving technology. There's already a lot of mainstream vehicles with advanced lane-keeping features and similar that are doing part of the driving autonomously, Tesla's "self-driving" features are pretty advanced, and we've had some autonomous trains and airport shuttlepods for a while now.
If you're asking how that sort of thing works - it's all about sensors like cameras and LIDAR to "see" what's going on around the vehicle and software to respond to what it's measuring in the appropriate way. A lot of regular human-driven cars are now "drive-by-wire" - controls like the throttle and brake aren't directly connected to a physical throttle or brake pad any more, they're sensors and switches that feed your input into a computer, and the computer sends out control actions to motors and solenoids that make the actual moving parts of the car go. For self driving cars, you simply add in another computer sending the inputs digitally based on sensors and cameras rather than a human pressing buttons and switches.
As to a sentient machine like the stories, that sort of thing is a long way off. We don't have any artificial general intelligence that's anywhere close to sentience. There's some pretty cool AI research like GPT-3 and neural networks, but it's still got a long way to go. Even if we were close, an AI clever enough to be sentient would be capable of a lot more than just being a driver - it starts bringing up questions of whether it would want to just be driving around, and what sort of rights an AI might have.
Also worth having a look at the answers on
https://www.careervillage.org/questions/41246/how-can-someone-get-into-building-autonomous-cars
and
https://www.careervillage.org/questions/45647/how-will-engineers-overcome-the-struggles-with-autonomous-vehichles
Joseph Neilson
Stefan Aubertin
General Manager
2
Answers
San Francisco, California
Updated
Stefan’s Answer
Self-driving cars, in particular robo-taxis, generally use a combination of technology to drive themselves which is quite complex and some manufacters of this technology differ in their approach.
Here's my best attempt at a short answer for a technology that is rapidly evolving. At present, self driving cars operate with an on-board computer running a stack of code and machine learning. The car will have several sensors used to "percieve" the environment in order to make decisions as to how to navigate the road. These typically include lidar, radar, cameras and sometimes audio. Lidar allows these vehicles to "see" things a human driver cannot by emitting a beam of light to read the environment. For example, lidar allows the car to see well around a corner beyond the field of human vision. Radar allows the car to detect items (cars, trucks, motorcycles, humans, etc) over a much longer distance allowing advanced information as to what lies on the road ahead. Cameras placed in various positions offer a short range visual input as to what exists in the immediate surroundings. Self-driving cars use all this, for what we call in the industry, perception: that is to percieve all of the elements that share the road. In the case of robo-taxis, this all works in concert with a pretty robust data set of accurate maps for the domain in which these cars operate. Providers of these services will have to perpetually update their maps to reflect any changes. Lastly, for outlier cases and special circumstances, most of these fleets are monitored remotely and have some infrastructure for what is called tele-operations. This allows a remote operator to navigate manually around special conditions.
Here's my best attempt at a short answer for a technology that is rapidly evolving. At present, self driving cars operate with an on-board computer running a stack of code and machine learning. The car will have several sensors used to "percieve" the environment in order to make decisions as to how to navigate the road. These typically include lidar, radar, cameras and sometimes audio. Lidar allows these vehicles to "see" things a human driver cannot by emitting a beam of light to read the environment. For example, lidar allows the car to see well around a corner beyond the field of human vision. Radar allows the car to detect items (cars, trucks, motorcycles, humans, etc) over a much longer distance allowing advanced information as to what lies on the road ahead. Cameras placed in various positions offer a short range visual input as to what exists in the immediate surroundings. Self-driving cars use all this, for what we call in the industry, perception: that is to percieve all of the elements that share the road. In the case of robo-taxis, this all works in concert with a pretty robust data set of accurate maps for the domain in which these cars operate. Providers of these services will have to perpetually update their maps to reflect any changes. Lastly, for outlier cases and special circumstances, most of these fleets are monitored remotely and have some infrastructure for what is called tele-operations. This allows a remote operator to navigate manually around special conditions.