Intention-Aware Guardian Architecture for Vehicle Overtaking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for supervising overtaking maneuvers in autonomous vehicles are conservative due to uncertainty in the intentions of other drivers, leading to smaller invariant sets and less permissive driving scenarios, which can result in reduced safety and increased conservatism.
Innovation Solution
A guardian architecture that includes an intention estimation module and a library of robust controlled invariant sets (RCIS) for different driver intention models, allowing for online estimation of the lead vehicle's intention and switching to a more permissive supervisor based on the estimated intention, ensuring safe and less conservative driving inputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If formal verification methods with set invariance are used to guarantee safety, then safety is improved, but the invariant sets become smaller and driving scenarios become more conservative
Solution Approach 1:
The system dynamically switches between different invariant sets based on the estimated intention of the lead vehicle. When the lead vehicle is estimated to be cautious, a more permissive invariant set is activated; when the lead vehicle is estimated to be aggressive, a more conservative invariant set is activated. This dynamic adaptation resolves the contradiction by making the safety constraints flexible rather than fixed.
Solution Approach 2:
The system changes the parameters of the invariant sets based on the lead vehicle's intention model. Different intention models (aggressive, cautious, cooperative) correspond to different parameter configurations of the invariant sets, allowing the system to adjust the degree of conservativeness to match the actual driving situation, thereby improving both safety and permissiveness.
2Reliability
If robust controlled invariant sets are used to account for model uncertainty and external disturbances, then reliability is improved, but the system becomes more conservative and less permissive
Solution Approach 1:
The system dynamically adjusts the level of robustness required based on the estimated intention of the lead vehicle. When the lead vehicle is estimated to be cooperative or cautious, the system relaxes the robustness requirements and allows more permissive driving inputs. When the lead vehicle is estimated to be aggressive, the system tightens the robustness requirements. This dynamic adjustment resolves the contradiction between robustness and permissiveness.
Solution Approach 2:
The system changes the parameters of the robust controlled invariant sets based on the intention estimation. Different intention models correspond to different parameter configurations that balance robustness and permissiveness. This allows the system to maintain safety robustness while avoiding unnecessary conservativeness in situations where the lead vehicle behavior is predictable and safe.
3Reliability
If a single conservative supervisor is used to handle all possible lead vehicle intentions, then safety is maintained, but the driving behavior becomes overly restrictive
Solution Approach 1:
The system segments the single supervisor into multiple specialized supervisors, each designed for a specific lead vehicle intention model (aggressive, cautious, cooperative). Each supervisor is optimized for its specific intention type, allowing for more flexible and less conservative driving behavior tailored to each scenario. The intention estimation module selects which supervisor to activate based on the current situation.
Solution Approach 2:
The system dynamically selects and switches between multiple supervisors based on the estimated intention of the lead vehicle. Instead of using a single conservative supervisor for all situations, the system activates the appropriate supervisor that matches the current driving scenario, enabling adaptive behavior that is neither overly conservative nor unsafe.
4Adaptability or versatility
If intention estimation is implemented to select appropriate intention models, then adaptability is improved, but system complexity increases
Solution Approach 1:
The system introduces an intention estimation module as an intermediary between the sensor data and the supervisor selection. This intermediary processes the sensor data and generates an estimated intention that guides the selection of the appropriate supervisor and invariant set. While this adds a component, it simplifies the overall architecture by providing a systematic way to adapt to different driving scenarios without requiring complex direct coupling between all possible intention models and supervisors.
Data Source
AI summary
In one aspect, the present disclosure provides a method in a data processing system comprising at least one processor and at least one memory, the at least one memory comprising instructions executed by the at least one processor to implement a vehicle overtaking monitoring system. The method comprises receiving, from a first plurality of sensors coupled to an ego vehicle, lead vehicle data about a lead vehicle, inferring an estimated intention of the lead vehicle based on the lead vehicle data, selecting an intention model from a plurality of intention models based on the estimated intention, calculating a set of permissible driving inputs of the ego vehicle based on the intention model, calculating at least one driver input range based on the set of permissible driving inputs, and causing the at least one driver input range to be displayed to a driver of the ego vehicle.