Hybrid Motion Planning for Human-Like Multi-Lane Trajectories
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous vehicle motion planning technologies struggle to replicate human driving behaviors accurately, especially in complex scenarios, and often prioritize short-term forecasting over long-term comfort and safety.
Innovation Solution
A hybrid motion planner system that combines rule-based and learning-based methods using a multi-lane intelligent driver model (MIDM) and a multi-lane hybrid planning driver model (MPDM), trained with open-loop ground truth data and close-loop simulations, to generate human-like accurate, comfortable, and safe trajectories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If rule-based motion planning methods are used, then short-term forecasting accuracy is improved, but long-term comfort and safety deteriorate
Solution Approach 1:
The patent combines rule-based methods (Intelligent Driver Model for short-term predictions) with learning-based methods (Deep Reinforcement Learning for long-term planning) into a hybrid motion planner. The rule-based component handles immediate safety and collision avoidance, while the learning-based component optimizes for long-term comfort and safety, resolving the contradiction between short-term accuracy and long-term reliability
Solution Approach 2:
The patent introduces a trajectory optimization module as an intermediary that refines the raw predictions from the Intelligent Driver Model. This optimizer adjusts trajectories to satisfy long-term comfort and safety constraints while maintaining the short-term forecasting accuracy of the rule-based component, acting as a mediator between the two conflicting requirements
2Reliability
If learning-based motion planning methods are used, then long-term comfort and safety are improved, but short-term forecasting accuracy deteriorates
Solution Approach 1:
The hybrid architecture merges the strengths of both approaches: the Intelligent Driver Model provides accurate short-term predictions based on physical laws, while the Deep Reinforcement Learning component enhances long-term comfort and safety. Neither approach alone is used, avoiding their respective weaknesses
3Measurement precision
If hybrid motion planning combining rule-based and learning-based methods is used, then motion planning accuracy is improved, but system complexity increases
Solution Approach 1:
The hybrid motion planner is segmented into distinct functional modules: the Intelligent Driver Model for short-term rule-based predictions, the Deep Reinforcement Learning component for long-term optimization, and a trajectory optimization module for refinement. This segmentation allows each component to be developed and tuned independently, managing system complexity while achieving high accuracy
Data Source
AI summary
Systems and methods for a hybrid motion planner for autonomous vehicles. A multi-lane intelligent driver model (MIDM) can predict trajectory predictions from collected data by considering adjacent lanes of an ego vehicle. A multi-lane hybrid planning driver model (MPDM) can be trained using open-loop ground truth data and close-loop simulations to obtain a trained MPDM. The trained MPDM can predict planned trajectories with collected data and the trajectory predictions to generate final trajectories for the autonomous vehicles. The final trajectories can be employed to control the autonomous vehicles.