Autonomous Lane Change Control With Gap Evaluation and DRL
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane change technologies for autonomous vehicles fail to flexibly handle collision situations, often resulting in accidents, and exhibit a tradeoff between lane change performance and safety.
Innovation Solution
A lane change system for autonomous vehicles that divides the lane change process into multiple stages, using a gap search and evaluation algorithm to analyze ease and safety, and employs a deep reinforcement learning algorithm to optimize lane changes, ensuring safety and performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing lane change technologies prioritize vehicle safety, then safety is improved, but lane change performance deteriorates
Solution Approach 1:
The lane change process is divided into multiple distinct stages: gap search stage, gap evaluation stage, and lane change execution stage. Each stage has specific sub-stages (e.g., ready stage, approaching stage, negotiation stage, lane change stage, completion stage) with dedicated transition conditions. This segmentation allows the system to optimize safety checks in gap search and evaluation while enabling decisive action in execution stages, resolving the contradiction between safety and performance.
Solution Approach 2:
The system dynamically adjusts transition conditions between stages based on real-time gap evaluations. The gap search and evaluation algorithm continuously assesses surrounding vehicle movements and adjusts the readiness criteria for lane change transitions. This dynamic adaptation allows the system to be conservative when safety risks are high and aggressive when opportunities are favorable, simultaneously improving safety and lane change performance.
2Productivity
If existing lane change technologies frequently perform lane changes, then lane change performance is improved, but safety deteriorates
Solution Approach 1:
The system performs preliminary gap search and evaluation before executing lane changes. The gap search algorithm predicts future movements of surrounding vehicles and identifies suitable gaps in advance. The evaluation algorithm assesses the safety and feasibility of each identified gap before the vehicle commits to a lane change maneuver. This preliminary action ensures that frequent lane changes are performed only after thorough safety verification, maintaining both high lane change performance and vehicle safety.
3Reliability
If existing lane change technologies do not perform lane changes, then safety is improved, but lane change performance deteriorates
Solution Approach 1:
The system continuously monitors surrounding vehicle positions, speeds, and predicted movements, providing real-time feedback to the gap search and evaluation algorithms. This feedback loop allows the system to identify optimal lane change opportunities that maximize safety while enabling necessary lane changes. The negotiation stage specifically uses feedback from surrounding vehicles to determine when it is safe to proceed, resolving the contradiction between being overly conservative and missing necessary lane changes.
Data Source
AI summary
Disclosed is a lane change system for an autonomous vehicle. A lane change method performed by a lane change system may include receiving driving state information of a vehicle and outputting a path and velocity plan corresponding to the input driving state information of the vehicle by using a lane change algorithm. The driving state information of the vehicle may include state information of the vehicle including driving information of the vehicle and situation information for recognizing a situation of a surrounding vehicle on the basis of the vehicle.