I've been designing and coding a self driving RC car inspired by Steven Gong's videos where he did the same. I've finally reached a milestone where the car can fully autonomously drive and...
I've been designing and coding a self driving RC car inspired by Steven Gong's videos where he did the same. I've finally reached a milestone where the car can fully autonomously drive and navigate! I hope to compete in the next F1Tenth competition.
The chassis I designed in fusion 360 and all of the code is running on a raspberry pi.
One of my next goals is to perform controlled drifting autonomously!
I am using a relatively inexpensive 2D LiDAR scanner (FHL-LD19). But because its a 2D scanner, all of the information I get is from an imaginary plane roughly 10cm above the ground. So the car...
I am using a relatively inexpensive 2D LiDAR scanner (FHL-LD19). But because its a 2D scanner, all of the information I get is from an imaginary plane roughly 10cm above the ground. So the car actually cant "see" the ground, just the walls around it.
So in theory it could navigate an overpass without hitting the walls, but it would have no idea where the lanes are in order to stay in them since they are just markings on the ground. I would need to add a camera to do lane following.
You can see the "ground" getting colored grey even though the car technically cant sense it. Its really just an assumption based on the fact that the LiDAR beam made it to a wall, so there must be nothing else in the way.
edit: I think I misunderstood your question initially. If you're asking whether or not my map structure could handle tunnels or bridges above/below an already explored area, it cannot. Since my sensor is a 2D scanner I developed the occupancy grid without elevation in mind unfortunately :(
That sounds like an interesting challenge though that I might try to implement later!
From what I saw on YouTube about those competitions there's enough work to be done just to be competitive, but the kind of system that would track motion well enough to drift would probably be...
One of my next goals is to perform controlled drifting autonomously!
From what I saw on YouTube about those competitions there's enough work to be done just to be competitive, but the kind of system that would track motion well enough to drift would probably be really helpful in competition just to handle unintentional slides.
For sure! It is definitely going to be quite the challenge, but if I can pull it off it would be very helpful for operating the car at its limit. Plus it would be quite the flex to show off...
For sure! It is definitely going to be quite the challenge, but if I can pull it off it would be very helpful for operating the car at its limit. Plus it would be quite the flex to show off drifting just for style!
Luckily I found a amazing Stanford paper that goes incredibly deep into autonomous drifting. The paper mainly focuses on the safety advantages in being able to still control a car that is sliding or has "saturated" tires. This paper will be invaluable for when I get to implementing a drift controller.
I've been designing and coding a self driving RC car inspired by Steven Gong's videos where he did the same. I've finally reached a milestone where the car can fully autonomously drive and navigate! I hope to compete in the next F1Tenth competition.
The chassis I designed in fusion 360 and all of the code is running on a raspberry pi.
One of my next goals is to perform controlled drifting autonomously!
Can I ask what you used for slam mapping? Can the car navigate an environment with overlapping lanes (like an overpass)?
I am using a relatively inexpensive 2D LiDAR scanner (FHL-LD19). But because its a 2D scanner, all of the information I get is from an imaginary plane roughly 10cm above the ground. So the car actually cant "see" the ground, just the walls around it.
So in theory it could navigate an overpass without hitting the walls, but it would have no idea where the lanes are in order to stay in them since they are just markings on the ground. I would need to add a camera to do lane following.
Here's a link to a video showing the raw LiDAR data getting processed into an occupancy grid
You can see the "ground" getting colored grey even though the car technically cant sense it. Its really just an assumption based on the fact that the LiDAR beam made it to a wall, so there must be nothing else in the way.
edit: I think I misunderstood your question initially. If you're asking whether or not my map structure could handle tunnels or bridges above/below an already explored area, it cannot. Since my sensor is a 2D scanner I developed the occupancy grid without elevation in mind unfortunately :(
That sounds like an interesting challenge though that I might try to implement later!
From what I saw on YouTube about those competitions there's enough work to be done just to be competitive, but the kind of system that would track motion well enough to drift would probably be really helpful in competition just to handle unintentional slides.
For sure! It is definitely going to be quite the challenge, but if I can pull it off it would be very helpful for operating the car at its limit. Plus it would be quite the flex to show off drifting just for style!
Luckily I found a amazing Stanford paper that goes incredibly deep into autonomous drifting. The paper mainly focuses on the safety advantages in being able to still control a car that is sliding or has "saturated" tires. This paper will be invaluable for when I get to implementing a drift controller.