Autonomous Vehicle Surprise Assessment for Predictable Trajectories
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles often behave differently from human drivers, leading to erroneous assumptions by other road users, which can result in uncomfortable and dangerous situations, such as sudden braking for non-existent objects, causing road users to adopt inappropriate following distances or take their eyes off the vehicle.
Innovation Solution
A method for controlling autonomous vehicles by generating probability distributions using a generative model to predict behaviors and assess surprise, comparing these distributions to determine the unexpectedness of the vehicle's trajectory to observers, and adjusting vehicle control accordingly to minimize surprise and prevent dangerous situations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If autonomous vehicles use non-human-like behavior patterns, then operational efficiency and task completion improve, but road user safety and predictability worsen
Solution Approach 1:
The system applies different behavior strategies to different aspects of driving: non-human-like efficient behavior for task execution, but human-like predictable behavior for safety-critical maneuvers. This local differentiation resolves the contradiction by allowing efficiency improvements without compromising safety.
Solution Approach 2:
Instead of making autonomous vehicles fully human-like or fully machine-optimized, the system inverts the approach by using machine-optimized behavior for efficiency while deliberately incorporating human-like predictability for safety, achieving both goals through opposite strategic applications.
2Adaptability or versatility
If autonomous vehicles exhibit unexpected behaviors, then adaptability to complex situations improves, but road user confusion and dangerous assumptions worsen
Solution Approach 1:
The system performs preliminary actions by communicating intent through predictable human-like behaviors before executing complex adaptive maneuvers. This allows road users to form correct assumptions about vehicle intentions, preventing confusion while maintaining adaptability.
Solution Approach 2:
Human-like predictable behavior acts as an intermediary between the autonomous vehicle's complex decision-making system and other road users. This mediator translates machine logic into human-understandable patterns, reducing confusion while preserving adaptability.
3Productivity
If autonomous vehicles optimize for task completion speed, then productivity improves, but interaction safety with vulnerable road users worsens
Solution Approach 1:
The system segments driving behaviors into safety-critical components (where human-like predictability is maintained) and efficiency-optimized components (where machine-optimized speed is applied). This segmentation allows simultaneous optimization of both productivity and safety by applying different strategies to different behavioral aspects.
Data Source
AI summary
Aspects of the disclosure provide for controlling an autonomous vehicle. For instance, a first probability distribution may be generated for the vehicle at a first future point in time using a generative model for predicting expected behaviors of objects and a set of characteristics for the vehicle at an initial time expected to be perceived by an observer. Planning system software of the vehicle may be used to generate a trajectory for the vehicle to follow. A second probability distribution may be generated for a second future point in time using the generative model based on the trajectory and a set of characteristics for the vehicle at the first future point expected to be perceived by the observer. A surprise assessment may be generated by comparing the first probability distribution to the second probability distribution. The vehicle may be controlled based on the surprise assessment.