Parallel Autonomy Vehicle Control via Neural Embedding Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Advanced driver-assistance systems (ADAS) in vehicles sometimes cause discomfort for drivers due to unexpected interventions, as the systems take control of steering, acceleration, and braking, leading to a lack of driver control perception.
Innovation Solution
A parallel autonomy system that uses neural networks to encode sensor data into an intermediate embedding space, processes it with both autonomous-behavior and driver-behavior models, and combines their outputs to produce an ideal behavior for vehicle control, balancing comfort and safety by incorporating driver intentions and minimizing risk.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the ADAS takes control of the vehicle to prevent hazardous outcomes, then safety is improved, but driver comfort and control perception deteriorate
Solution Approach 1:
The system merges the autonomous vehicle's planned behavior with the driver's intended behavior into a fused ideal behavior. The ideal-behavior model combines outputs from both the autonomous-behavior model and driver-behavior model, integrating safety-oriented autonomous control with driver comfort and intention, thereby resolving the contradiction between safety intervention and driver comfort.
Solution Approach 2:
The ideal-behavior model serves as an intermediary that mediates between the autonomous system's safety-critical interventions and the driver's intended actions. It processes both the planned behavior from the autonomous system and the predicted behavior from the driver-behavior model to generate a balanced ideal behavior that maintains safety while respecting driver comfort and control perception.
2Reliability
If the ADAS frequently intervenes to ensure safety, then incident risk is reduced, but driver control perception and comfort worsen
Solution Approach 1:
The system applies partial action by selectively integrating only the necessary safety-critical components of the autonomous planned behavior with the driver's intended behavior. Rather than fully overriding the driver, the ideal-behavior model combines both behaviors, applying autonomous intervention only to the extent needed to reduce incident risk while preserving driver control perception for non-critical aspects.
3Reliability
If the system fully overrides driver control to maximize safety, then incident prevention is improved, but driver comfort and acceptance deteriorate
Solution Approach 1:
The system dynamically adjusts the balance between autonomous planned behavior and driver predicted behavior based on the situation. The ideal-behavior model continuously fuses both behavior outputs, allowing the degree of autonomous intervention to vary dynamically according to the driving context, thereby maintaining both incident prevention and driver acceptance.
Data Source
AI summary
Systems and methods for parallel autonomy of a vehicle are disclosed herein. One embodiment receives input data, the input data including at least one of sensor data and structured input data; encodes the input data into an intermediate embedding space using a first neural network; inputs the intermediate embedding space to a first behavior model and a second behavior model, the first behavior model producing a first behavior output, the second behavior model producing a second behavior output; combines the first behavior output and the second behavior output using an ideal-behavior model to produce an ideal behavior for the vehicle; and controls one or more aspects of operation of the vehicle based, at least in part, on the ideal behavior.