Competence-Aware Autonomy Control With Online State Space Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in determining the appropriate level of operational autonomy due to insufficient domain modeling and reliance on human intervention, which can lead to inefficient and costly operations.
Innovation Solution
The implementation of a competence-aware system (CAS) that uses an autonomy cognizant agent (ACA) to select actions and autonomy levels based on an autonomy model and feedback model, allowing for iterative state space refinement and proactive planning to minimize human intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If autonomous vehicles rely on human intervention to determine operational autonomy levels, then safety can be maintained through human oversight, but operational efficiency decreases and costs increase
Solution Approach 1:
The system enables autonomous vehicles to self-determine their operational autonomy levels through the ACA and competence model, eliminating the need for continuous human intervention. The vehicle independently evaluates its own competence across different driving scenarios and autonomously selects appropriate autonomy levels, thereby maintaining safety through self-assessment while improving operational efficiency by removing human oversight bottlenecks.
Solution Approach 2:
The system implements a feedback mechanism where the ACA continuously monitors vehicle performance, sensor data, and environmental conditions to dynamically adjust autonomy levels. Human feedback is incorporated during training phases to refine the competence model, creating a closed-loop system that maintains safety through continuous learning and adaptation while enabling autonomous operation during normal conditions.
2Measurement precision
If autonomous vehicles use comprehensive domain modeling to determine autonomy levels, then operational accuracy improves, but system complexity increases
Solution Approach 1:
The system segments the complex driving environment into distinct operational scenarios (e.g., highway driving, urban navigation, adverse weather) and evaluates competence separately for each scenario. The ACA maintains a competence model that assesses autonomy capability across multiple discrete dimensions rather than as a single monolithic assessment, reducing overall system complexity while maintaining comprehensive evaluation accuracy.
Solution Approach 2:
The system applies different levels of modeling detail to different operational contexts. The ACA evaluates competence with high precision for critical safety scenarios while using coarser assessments for less critical operations. This localized approach to modeling precision reduces computational complexity while maintaining operational accuracy where it matters most.
3Productivity
If autonomous vehicles minimize human intervention through proactive planning, then productivity increases, but reliability may decrease due to reduced human oversight
Solution Approach 1:
The ACA performs preliminary assessment of operational autonomy levels before executing driving tasks. By proactively planning and determining the appropriate autonomy level in advance based on predicted environmental conditions and vehicle competence, the system enables continuous autonomous operation without waiting for human decisions, thereby increasing productivity while maintaining reliability through pre-validation of autonomy level selections.
Data Source
AI summary
A first method includes detecting, based on sensor data, an environment state; selecting an action based on the environment state; determining an autonomy level associated with the environment state and the action; and performing the action according to the autonomy level. The autonomy level can be selected based at least on an autonomy model and a feedback model. A second method includes calculating, by solving an extended Stochastic Shortest Path (SSP) problem, a policy for solving a task. The policy can map environment states and autonomy levels to actions and autonomy levels. Calculating the policy can include generating plans that operate across multiple levels of autonomy.