Autonomous Driving Path Control Using Occupancy Grid Yaw Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous driving systems rely on end-to-end approaches that are difficult to integrate into vehicles other than those they were designed for, require high-frequency accurate position determination, and lack human-like driving behavior, leading to inefficiencies and increased computational costs.
Innovation Solution
A modular system that determines vehicle traveling data using a preliminary path with predefined coordinates and speed limits, combined with an occupancy grid and neural networks to predict yaw and velocity values, allowing for flexible integration and human-like driving behavior without relying on high-frequency accurate position determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If end-to-end AI systems are used for autonomous driving, then driving commands can be generated automatically, but the system becomes difficult to integrate into vehicles other than those it was designed for
Solution Approach 1:
The system divides the autonomous driving task into separate modules: a path planning module that generates reference paths and a control module that executes commands. This segmentation allows each module to be independently trained and integrated, improving vehicle compatibility while maintaining automation.
Solution Approach 2:
The path planning module generates reference paths that can be applied to different vehicle types and scenarios. By creating a universal path representation that is vehicle-agnostic, the system enables easy integration across multiple platforms without retraining the entire system.
2Reliability
If high-frequency accurate position determination is used, then reliable travelling paths can be determined, but computational costs and system complexity increase
Solution Approach 1:
The system pre-generates reference paths based on map data and destination information before actual vehicle operation. This preliminary path planning reduces the need for real-time position calculations, lowering computational requirements while maintaining reliability.
Solution Approach 2:
The reference path acts as an intermediary between the destination and the vehicle's actual position. Instead of continuously calculating optimal paths based on real-time position data, the system follows pre-computed reference paths, reducing positioning system complexity.
3Productivity
If visual sensor input only is used, then the system can process information efficiently, but it lacks human-like driving behavior adaptation
Solution Approach 1:
The system incorporates multiple input parameters beyond visual data, including vehicle state information (speed, acceleration, steering angle) and contextual factors. By changing the parameter set from single-source visual input to multi-source sensor fusion, the system achieves both efficiency and human-like adaptability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Apparatus 20 for determining vehicle travelling data for an autonomous driving control system 1 of a vehicle 100 comprising: an input unit configured to receive an occupancy grid ocg from an occupancy grid generation unit 300 of the autonomous driving control system 1 and a preliminary travelling path from a preliminary travelling path generation unit 200 of the autonomous driving control system 1, wherein the preliminary travelling path includes coordinates X, Y for a plurality of positions along the preliminary travelling path; a determination unit configured to determine yaw and velocity values of future travelling positions of the vehicle based on the occupancy grid and the preliminary travelling path received by the input unit; and an output unit which outputs the determined yaw and velocity values of the vehicle 100.