Learning-Based Speed Planner for Autonomous Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speed planning methods for autonomous driving vehicles are computationally expensive and fail to mimic human driving behaviors effectively, making them inefficient for navigating around obstacles.
Innovation Solution
A learning-based speed planning system that collects human driving data, converts it into feature vectors, and uses machine learning to train speed models, allowing autonomous vehicles to navigate obstacles by mimicking human driving decisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If dynamic programming or quadratic programming is used to optimize speed, then speed optimization is achieved, but computational cost becomes very expensive
Solution Approach 1:
The patent replaces traditional mechanical optimization methods (dynamic programming, quadratic programming) with a machine learning-based speed planner. The system trains a neural network model offline to learn optimal speed patterns from human driving data, then uses this pre-trained model for real-time speed planning. This substitution transforms the computationally intensive online optimization problem into a lightweight inference task, dramatically reducing computational cost while maintaining speed optimization capability.
2Reliability
If traditional speed planning methods are used, then speed optimization is achieved, but ability to match human driving behaviors is difficult
Solution Approach 1:
The patent directly copies human driving behaviors by collecting speed data from human drivers and using it to train the machine learning model. The system records human drivers' speed choices in various driving scenarios and uses this empirical data to teach the AI system how humans actually drive. This copying approach enables the autonomous vehicle to replicate natural human driving patterns, making the speed planning more adaptable to human expectations and behaviors.
Solution Approach 2:
The patent changes the fundamental parameter of how speed decisions are made - from rule-based or optimization-based parameters to learning-based parameters. By using machine learning, the system can capture complex, non-linear relationships between driving conditions and speed choices that traditional methods cannot represent. The model learns optimal speed parameters directly from data, enabling it to adapt to various human driving styles and scenarios.
Data Source
AI summary
A learning based speed planner for autonomous driving vehicles (ADV) is disclosed. An ADV is set into human-driving mode. Driving control elements are under control of a human driver, and other ADV logic is enabled. The ADV plans a route path on a segment of the route having an obstacle. ADV logic generates a station-time graph for the path of the segment, and a grid of cells to encompass the path and obstacle. A feature vector is generated from the grid. Human driving behavior is recorded as the ADV is navigated along the path. Recorded driving data for a large plurality of paths, obstacles and ADVs is transmitted to a server to generate a speed model. The speed model is downloaded to one or more ADVs for use in autonomous driving mode, to determine an initial speed to use in similar driving situations.