Autonomous Vehicle Task Assignment Under Imperfect Sensing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing task scheduling algorithms for autonomous vehicles fail to account for imperfect sensing conditions due to manufacturing defects or aging of sensors, leading to ineffective execution of traversal tasks.
Innovation Solution
An analytical model that estimates expected travel duration and collision probability under imperfect sensing conditions, using a statistical error model to define zone types and optimize task assignment based on completion time and collision count.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If perfect sensing conditions are assumed in task scheduling, then optimization algorithms can be simplified, but the reliability of task execution deteriorates under actual imperfect sensing conditions
Solution Approach 1:
The patent applies preliminary action by pre-characterizing sensor errors through statistical error models before task execution. The system collects sensor data under known conditions, establishes error distributions, and uses these pre-computed models to adjust task assignments in real-time, rather than reacting to errors after they occur during task execution
Solution Approach 2:
The patent changes parameters by incorporating sensor error statistics into the task scheduling algorithm. Instead of using deterministic sensor readings, the system uses probabilistic parameters (mean and standard deviation of sensor errors) to compute expected completion times and collision counts, transforming the scheduling problem from deterministic to stochastic optimization
2Device complexity
If sensor error is not characterized, then system design is simpler, but the accuracy of completion time and collision count estimation deteriorates
Solution Approach 1:
The system applies self-service by having the autonomous vehicles themselves generate the error characterization data during normal operation. The vehicles collect sensor readings under known ground truth conditions and automatically build their own statistical error models, eliminating the need for external calibration equipment or manual error characterization
Solution Approach 2:
The patent implements feedback by using the computed statistical error models to continuously refine task assignments. The system monitors actual sensor performance, updates error statistics, and feeds this information back into the scheduling algorithm to improve future task assignments, creating a closed-loop system that learns from experience
3Productivity
If traditional optimized assignment algorithms are used without sensor error consideration, then computational efficiency is maintained, but the effectiveness of task execution deteriorates under imperfect sensing
Solution Approach 1:
The patent applies dynamics by making the task assignment algorithm adaptive to changing sensor conditions. Instead of static optimization based on nominal sensor performance, the system dynamically adjusts assignments based on real-time sensor error statistics, allowing the scheduling policy to evolve as sensor characteristics change due to aging, environmental factors, or degradation
Data Source
AI summary
The disclosure relates generally to methods and systems for optimized assignment of traversal tasks under imperfect sensing of autonomous vehicles. Most of the techniques assumes perfect equipment conditions. With the imperfect sensing, most of the optimized assignment and scheduling algorithm may not be effective during actual execution of the tasks. The present disclosure solves the technical problems in the art by providing an analytical model which estimates the basic performance metrics such as an expected travel duration and safety estimation such as collision probability on its path, under imperfect sensing, for optimal assignment of the tasks. An analytical model is integrated with a performance estimator as implemented by the systems of the present disclosure, which tracks, predicts, and alerts on any major deviations from its intended performance of safety parameters.