Autonomous Vehicle Tree Search for Degraded Road Trajectory Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in navigating accurately and efficiently under altered or degraded road conditions, such as obscured or invalid roadway indicators, and unpredictable environmental and human behaviors, leading to potential stutters or the need for remote operator intervention.
Innovation Solution
An autonomous vehicle guidance system employing a tree search technique that classifies dynamic objects as reactive or passive using machine-learned models, predicts future environmental states, and determines candidate actions to generate paths, reducing computational load and increasing navigation scenarios by alternating cost estimates and default control instructions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the autonomous vehicle uses traditional path planning methods relying on roadway indicators, then the navigation is simple and computationally efficient, but the vehicle fails to navigate accurately when road conditions are altered or degraded
Solution Approach 1:
The path planning is segmented into multiple candidate paths generated through tree search exploration, rather than relying on a single normative path based on roadway indicators. This allows the system to evaluate multiple possible trajectories and select the safest path even when road conditions are degraded or altered.
Solution Approach 2:
The system performs preliminary exploration of multiple candidate paths and predicts future environmental states before committing to a navigation decision. This advance planning allows the vehicle to prepare for altered road conditions and unpredictable scenarios before they require immediate response.
2Reliability
If the autonomous vehicle explores more candidate actions and paths using tree search, then the navigation safety and scenario coverage increase, but the computational load increases
Solution Approach 1:
The system performs partial exploration of the action space by limiting the tree search to a manageable depth and breadth, focusing on the most promising candidate paths. This allows sufficient safety verification without requiring exhaustive exploration of all possible trajectories, thus balancing computational load with navigation safety.
Solution Approach 2:
The system dynamically adjusts tree search parameters such as exploration depth, number of candidate paths, and simulation steps based on computational resources available and environmental complexity. This allows the vehicle to maintain high safety standards while adapting computational effort to match available processing power.
3Speed
If the autonomous vehicle uses normative path planning based on roadway indicators, then the computational process is fast, but the vehicle may stutter or require remote operator intervention when encountering unpredictable scenarios
Solution Approach 1:
The system performs periodic tree search evaluations at discrete time steps along the navigation path, rather than continuously re-planning. This periodic re-evaluation of candidate paths maintains computational speed while ensuring the vehicle can respond to unpredictable scenarios and environmental changes without excessive computational overhead.
4Measurement precision
If the autonomous vehicle classifies dynamic objects as reactive or passive using machine-learned models, then the prediction of future environmental states improves, but the system complexity increases
Solution Approach 1:
The system extracts only the essential classification feature (reactive vs. passive) from dynamic objects using machine-learned models, rather than analyzing all possible object attributes. This extraction of the most critical distinction improves prediction accuracy for future environmental states while minimizing the added system complexity.
Data Source
AI summary
Trajectory generation for controlling motion or other behavior of an autonomous vehicle may include alternately determining a candidate action and predicting a future state based on that candidate action. The technique may include determining a cost associated with the candidate action that may include an estimation of a transition cost from a current or former state to a next state of the vehicle. This cost estimate may be a lower bound cost or an upper bound cost and the tree search may alternately apply the lower bound cost or upper bound cost exclusively or according to a ratio or changing ratio. The prediction of the future state may be based at least in part on a machine-learned model's classification of a dynamic object as being a reactive object or a passive object, which may change how the dynamic object is modeled for the prediction.