I'm working on my graduation project with a classmate, where we plan to build an RC car controlled by a neural network. While we have some experience with self-driving RC cars and have done well in competitions, we often run into issues finding quality sensors. Previously, we've tried homemade IR sensors, which were inconsistent, and VL53L1X sensors that turned out to be too slow.
With our new approach using a neural network that requires precise and rapid data, we can't rely on our old sensors and software tricks anymore. I've looked into several sensor options:
- GARMIN LIDAR-Lite V3: Quite expensive and likely overkill.
- GARMIN LIDAR V4: Not true LIDAR and may behave differently under varying lighting conditions.
- Benewake TF-Luna: Its tolerance of ±6 cm is excessive for a track just 1m wide.
- Benewake TFmini-S: Faces the same tolerance issue as the TF-Luna.
- RPLIDAR A1: This one is too slow as well.
What I'm looking for is either a single sensor that can measure up to 8m, be quick enough for measurements every 20ms, and accurate to ±2.5cm, or a budget-friendly solution with multiple 1D sensors (up to five or eight) totaling around 350€. Any recommendations would be greatly appreciated!
2 Answers
Hey! Are you locked into using LIDAR? If not, you might want to look into RADAR sensors instead. I've had issues with the Garmin LIDAR V3 not reporting range accurately. I suggest checking out OmniPreSense RADAR Modules. They have a great detection range (1-25m), speed reporting, and decent prices. Plus, they’re well-documented and should interface well with Arduino or Raspberry Pi.
Regarding the VL53L1X sensors, you mentioned they were too slow. They actually have a 100Hz output that’s pretty responsive, so maybe it’s worth checking your code? It might be a small glitch that's slowing down your measurements.