Iterative Learning Control for LiDAR Memory and Convergence
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
LiDAR systems face challenges with high memory requirements and slow convergence times due to large error tables and iterative learning controller tables, particularly when reconfiguring scanning profiles, such as when a vehicle's pitch changes relative to the horizon.
Innovation Solution
Implementing iterative learning control methods that pre-learn feedforward tables and approximate error tables and iterative learning controller tables using piecewise functions, reducing memory needs and accelerating convergence by determining iterative learning functions based on reference profiles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If iterative learning control uses large error tables and iterative learning controller tables, then measurement precision and control accuracy are improved, but memory requirements increase and convergence time increases
Solution Approach 1:
The patent transforms discrete lookup tables into continuous piecewise function representations. Instead of storing numerous discrete error values and learning coefficients in large tables, the system uses mathematical functions with parameters that can be evaluated continuously, significantly reducing memory requirements while maintaining control accuracy.
Solution Approach 2:
The patent creates a simplified mathematical model (piecewise function) that copies the essential behavior of the complex lookup tables without retaining their full data structure. This functional copy captures the error correction characteristics while using minimal memory compared to the original tables.
2Measurement precision
If iterative learning control uses large error tables and iterative learning controller tables, then measurement precision and control accuracy are improved, but convergence time increases
Solution Approach 1:
By changing from discrete table lookups to continuous piecewise function evaluations, the system achieves faster convergence. The mathematical functions can be evaluated more efficiently than table lookups, and the continuous nature of the functions allows for smoother error correction and faster adaptation to changing conditions.
Solution Approach 2:
The patent pre-computes and stores the piecewise function parameters during system initialization or offline calibration. This preliminary action allows the system to use pre-processed mathematical models during real-time operation, reducing the computational burden and accelerating convergence compared to computing error corrections on-the-fly from large tables.
3Adaptability or versatility
If LiDAR system reconfigures scanning profiles (e.g., vehicle pitch changes), then adaptability is improved, but convergence time increases due to large error tables
Solution Approach 1:
The piecewise function representation allows for efficient reconfiguration by simply changing the function parameters or selecting different function segments rather than reloading large error tables. This enables rapid adaptation to new scanning profiles or vehicle attitudes with minimal convergence time.
Solution Approach 2:
The system dynamically adjusts the piecewise function parameters or selected functions based on the current operating conditions (e.g., vehicle pitch angle). This dynamic adaptation allows the system to quickly respond to changing requirements without the delay of processing large static tables, achieving fast convergence during reconfiguration.
Data Source
AI summary
A system includes a light emitter configured to emit light pulses. The system includes a controller configured to control positions of the emitted light pulses based on a selected reference profile. The controller includes an iterative learning controller configured to provide an error correction signal based on one or more iterative learning functions that correspond to the selected reference profile. The system includes a detector configured to provide one or more detection signals in response to a detection of a return pulse corresponding to at least one of the emitted light pulses.