AV Exception Playback for Faster Tele-Operator Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles often encounter situations that require human intervention due to limitations in their programming, leading to inefficiencies and a high demand for tele-operator assistance, which is not sustainable or cost-effective.
Innovation Solution
Implementing a system that allows autonomous vehicles to classify obstruction situations, assess risks, and autonomously navigate around them or request tele-operator assistance, reducing the need for extensive human intervention by storing and transmitting data snapshots for validation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If autonomous vehicles request tele-operator assistance for all exception situations, then the reliability of exception handling is improved, but the operational cost and demand for tele-operator support increase significantly
Solution Approach 1:
The system performs preliminary actions by capturing and storing playback data (snapshots of sensor data, system state, and environment) before tele-operator assistance is requested. This preliminary data collection enables tele-operators to efficiently analyze exception situations without needing to request additional real-time data, thereby improving exception handling reliability while reducing the time and resources required for tele-operator support.
Solution Approach 2:
The system creates copies of the autonomous vehicle's operational state through playback data, including snapshots of sensor data, system state, and environmental conditions. These copies allow tele-operators to review and analyze exception situations without interfering with the actual vehicle operation, improving reliability of exception handling while reducing the burden on tele-operator resources.
2Reliability
If the autonomous vehicle stores and transmits playback data for tele-operator review, then the quality of tele-operator decision-making is improved, but the loss of time for data preparation and transmission increases
Solution Approach 1:
The system performs preliminary data capture and organization during normal operation, storing playback data in an easily accessible format. When an exception occurs, the relevant data is already prepared and can be quickly transmitted to tele-operators, improving decision quality while minimizing the time loss associated with data preparation.
Solution Approach 2:
The system extracts only the relevant playback data necessary for tele-operator decision-making, rather than transmitting all possible data. By selectively extracting and transmitting only the critical snapshots and system states needed for exception analysis, the system improves decision quality while reducing data preparation and transmission time.
3Measurement precision
If the autonomous vehicle captures detailed snapshots at every time point, then the measurement precision of exception situation analysis is improved, but the quantity of data stored and transmitted increases
Solution Approach 1:
The system applies local quality by capturing detailed snapshots at critical moments during exception situations rather than uniformly at all times. By focusing high-resolution data collection on specific time points when exceptions occur or are likely to occur, the system achieves high measurement precision for exception analysis while minimizing the overall data volume stored and transmitted.
Solution Approach 2:
The system uses partial action by capturing snapshots at selected intervals and only when relevant to exception situations, rather than continuously capturing all data. This approach provides sufficient measurement precision for analyzing exception situations while significantly reducing the quantity of data that must be stored and transmitted compared to continuous full-resolution data capture.
Data Source
AI summary
Resolving an exception situation in autonomous driving includes receiving an assistance request to resolve the exception situation from an autonomous vehicle (AV); identifying a solution to the exception situation; forwarding the solution to a tele-operator; receiving a request for playback data from the tele-operator; receiving, from the AV, the playback data; and obtaining, from the tele-operator, a validated solution based on the tele-operator using the playback data. The playback data includes snapshots ni of data related to autonomous driving stored at the AV at respective consecutive times ti, for i=1, . . . , N.