Autonomous Driving Policy Personalization via Driver Preference Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computerized advanced driver assist and autonomous driving systems lack customization and are vulnerable to malicious attacks, as they often mimic a generic, risk-averse driving style and may respond uniformly in emergent situations, failing to account for individual driving preferences and local customs.
Innovation Solution
A system that receives and applies individual driving preferences to customize policies for computerized assist or autonomous driving, using data from sensors and machine learning algorithms to adapt driving styles to specific drivers or passengers, reducing the likelihood of uniform responses and enhancing security against malicious attacks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If computerized driver assist or autonomous driving systems mimic an idealized rule-following, risk-averse driving style, then safety and reliability are improved, but customization and adaptability to individual driving preferences deteriorate
Solution Approach 1:
The system dynamically adapts driving policies by learning individual driver preferences through machine learning algorithms. The driving behavior is not fixed but evolves based on collected data about the driver's preferences, creating a dynamic system that balances safety requirements with personalized driving styles.
Solution Approach 2:
The system changes key parameters of driving behavior by learning from sensor data and adjusting policy parameters to match individual driver preferences. This includes modifying acceleration patterns, braking styles, and response times while maintaining safety constraints, thereby transforming generic driving parameters into personalized ones.
2Ease of manufacture
If autonomous driving systems use generic algorithms, then ease of manufacture and deployment are improved, but vulnerability to malicious attacks and uniform responses in emergent situations worsen
Solution Approach 1:
The system applies local quality by customizing driving behavior to individual drivers rather than using a uniform approach. Each driver receives personalized driving policies based on their specific preferences and behaviors, making the system resistant to attacks that rely on predictable uniform responses while maintaining ease of deployment through automated learning.
3Reliability
If computerized driving systems provide risk-averse, rule-following behavior, then reliability and safety are improved, but the driving experience becomes generic and less engaging
Solution Approach 1:
The system uses feedback from sensors that continuously monitor driver behavior and preferences. This feedback loop allows the system to learn what the driver prefers and adjust the autonomous driving behavior accordingly, providing both safety through rule-following and an engaging experience through personalization.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Apparatuses, methods and storage medium associated with computerized assist or autonomous driving (CA/AD) of vehicles are disclosed herein. In various embodiments, an apparatus may include a CA/AD system to: receive an identifier identifying a driver/passenger of a vehicle; request or retrieve, using the identifier, individual driving preferences of the driver/passenger; and apply the individual driving preferences of the driver/passenger to policies for CA/AD of the vehicle, to customize the policies for CA/AD of the vehicle for the driver/passenger. The CA/AD system may further receive data for policy parameters of the customized policies; and CA/AD the vehicle, in a manner that is adapted for the individual, in accordance with the customized policies, based at least in part on the data for the policy parameters of the customized policies. Other embodiments may be described and claimed.