Clothoid commands in nucleotide board

Hi everyone,

I'm working on a robotics project using an STM32 Nucleo board for the BFMC robotics competition, and I'm trying to implement clothoid-based path planning for smooth trajectory generation. However, I'm facing issues with incorrect or non-exact command execution. The steering and velocity values sometimes deviate from the expected clothoid curve, likely due to numerical precision errors or control loop delays.

I've considered using fixed-point arithmetic for better precision and tuning a PID controller to correct deviations dynamically. Additionally, I’m thinking of applying a low-pass filter to sensor data to reduce noise affecting the control loop.

Has anyone else dealt with similar challenges? What strategies have worked for you to ensure accurate execution of clothoid commands on embedded systems? Any insights or suggestions would be greatly appreciated!

Thanks in advance!

Quote from SkyeFire

Speaking from 0 experience with the board or path planning technique you're using, but in general, filtering on sensor data is almost always required. How much depends on how much noise you're experiencing and how vulnerable your control setup is to said noise.

PID control can get pretty complex, but for realtime control, it's often a lifesaver. Tuning usually requires a lot of test runs and careful tweaking.

Sensor filtering is probably the lowest-hanging fruit to go after first.

Thanks a lot for the response. I have considered some of those and i will try and implement the rest of the proposals accordingly.

Quote from panic mode

there are many ways to filter data. one of the simplest is to average. one simple way to use averaging is to apply weights to new and old data.

X(n+1)=X(n)*T+ Xnew*(1-T)

where T is adjustable value for filtering (0...1). of course there are many others (butterworth, kalmann, fast kalman,...) it all depends on resources you have at your disposal.

Thanks, i really appreciate the recommendations. I will try many filters in order to find the best fit for the steers and manoeuvres needed.