Deep Learning based State Estimation

Simplified schema of a 1D CNN:

State Estimation

State estimation is the process of inferring the unobservable variables or "state" of a system based on observed data and a mathematical model of the system dynamics. It plays a crucial role in various fields such as control theory, robotics, and signal processing. Through techniques like Kalman filtering or particle filtering, state estimation enables predictions, control decisions, and system understanding even in the presence of uncertainty or noise. It's a fundamental tool for making informed decisions and achieving accurate control in complex systems.

Kinematic Bicycle Model[1]

The kinematic bicycle model is a simplified representation of a vehicle's motion, commonly used in robotics, control systems, and autonomous driving. It approximates the vehicle as a point mass with two wheels, allowing for straightforward analysis of its motion based on steering angle and velocity inputs. Despite its simplicity, the kinematic bicycle model provides valuable insights into vehicle dynamics and is a foundation for more complex vehicle models used in simulation and control algorithms.
The model can be defined with the following state-space representation,

$$ \frac{d}{dt} \begin{pmatrix} x \\ y \\ \theta \\ v \end{pmatrix} = \begin{pmatrix} v\cos{\theta} \\ v\sin{\theta} \\ \frac{v\tan{\delta}}{L} \\ a \end{pmatrix} $$

where $v$ is the vehicle's velocity in the x-axis, $\theta$ is the vehicle's yaw, $\delta$ is the steering angle, $L$ is the vehicle's wheelbase, $a$ is the acceleration/throttle, $f$ is friction in the x-axis.

References

[1] https://github.com/winstxnhdw/KinematicBicycleModel