Autonomous Vehicle Trajectory Simulation for Dynamic Obstacle Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous vehicle control systems fail to accurately predict collisions with dynamic obstacles and do not consider the interactions with other vehicles or the typical responses of human drivers, leading to suboptimal behavior in real-world scenarios.
Innovation Solution
A data-driven system and method for real-world autonomous vehicle trajectory simulation that collects vehicle sensor data to build models simulating human driving patterns, using sensors like cameras, LIDAR, and radar to generate trajectory prediction models that account for driver intentions and vicinal scenarios, enabling more accurate prediction of vehicle trajectories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional polynomial-based trajectory control is used, then the system is simple to implement, but it cannot accurately predict collisions with dynamic obstacles
Solution Approach 1:
The system segments the trajectory prediction problem into multiple polynomial segments (first trajectory portion and second trajectory portion) with different degrees. The first portion uses a lower-degree polynomial for simplicity, while the second portion uses a higher-degree polynomial for improved accuracy near the predicted endpoint, resolving the contradiction between computational simplicity and prediction accuracy.
Solution Approach 2:
The system dynamically adjusts the polynomial degree based on the prediction horizon and obstacle proximity. As the vehicle approaches the predicted endpoint or when dynamic obstacles are detected, the system transitions to higher-degree polynomials to improve collision prediction accuracy, while maintaining lower-degree polynomials for distant predictions to preserve computational efficiency.
2Reliability
If conventional control systems ignore human driver behavior, then the control logic is simpler, but the autonomous vehicle cannot achieve optimal behavior in real-world scenarios
Solution Approach 1:
The system creates simplified computational models that copy essential human driver behaviors, such as typical response times, acceleration patterns, and obstacle avoidance strategies. These behavioral models are integrated into the trajectory prediction to improve real-world performance without requiring full complexity of human decision-making processes.
Solution Approach 2:
The system incorporates human driver behavior parameters (e.g., reaction time distributions, acceleration preferences, following distances) into the polynomial trajectory generation. By adjusting these parameters based on observed human driving patterns, the autonomous vehicle achieves more natural and reliable behavior in real-world scenarios.
3Measurement precision
If the polynomial degree is increased to improve trajectory accuracy, then prediction precision improves, but computational load increases
Solution Approach 1:
The trajectory is divided into segments with different polynomial degrees. Distant trajectory portions use lower-degree polynomials requiring less computational power, while near-future portions use higher-degree polynomials for improved accuracy where precision is most critical for collision avoidance.
Solution Approach 2:
The system applies higher-degree polynomials only partially - specifically for the second trajectory portion near the predicted endpoint and when dynamic obstacles are present. For the majority of the trajectory (first portion), lower-degree polynomials are used, reducing overall computational load while maintaining sufficient accuracy for safety-critical regions.
Data Source
AI summary
A system and method for real world autonomous vehicle trajectory simulation may include: receiving training data from a data collection system; obtaining ground truth data corresponding to the training data; performing a training phase to train a plurality of trajectory prediction models; and performing a simulation or operational phase to generate a vicinal scenario for each simulated vehicle in an iteration of a simulation. Vicinal scenarios may correspond to different locations, traffic patterns, or environmental conditions being simulated. Vehicle intention data corresponding to a data representation of various types of simulated vehicle or driver intentions.