Tree Search Cost Estimation for Real-Time Vehicle Trajectory Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles struggle to navigate accurately and efficiently in non-standard driving conditions due to obscured, degraded, or invalidated roadway indicators, and unpredictable environmental factors, leading to increased computational intensity that is not feasible for real-time operation on consumer-grade hardware.
Innovation Solution
A machine-learned cost estimation architecture is integrated into a tree search technique to determine candidate actions and predict future states, reducing computational load by estimating costs instead of simulating them, allowing real-time operation on consumer-grade hardware.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing solutions for non-standard driving conditions are implemented, then navigation accuracy in obscured or degraded roadway conditions is improved, but computational intensity increases to the point of infeasibility for real-time operation on consumer-grade hardware
Solution Approach 1:
The patent replaces traditional simulation-based cost calculation (mechanical/computational system) with a machine-learned cost estimation model. The neural network architecture processes sensor data and predicts costs directly, substituting the computationally intensive simulation process with a trained model that provides similar accuracy at fraction of the computational cost, enabling real-time operation on consumer-grade hardware
Solution Approach 2:
The cost estimation architecture is trained in advance using simulated data to learn cost patterns without requiring real-time simulation. The pre-trained model performs cost estimation by directly processing current sensor inputs, eliminating the need for computationally expensive on-the-fly simulations during actual vehicle operation
2Productivity
If computational intensity is reduced for real-time operation, then operation feasibility on consumer-grade hardware is improved, but the number of navigable scenarios decreases
Solution Approach 1:
The machine-learned cost estimation architecture serves multiple functions: it handles standard driving conditions, obscured roadway indicators, degraded lane markings, invalidated indicators, and unpredictable environmental factors through a single unified model. This multi-functional approach maintains broad scenario coverage while operating efficiently on consumer-grade hardware in real-time
Data Source
AI summary
A machine-learned architecture for estimating the cost determined by a cost function for a prediction node of a tree search for exploring potential paths for controlling a vehicle may include two portions: a set up portion that includes models trained to process static data and a second portion that processes dynamic object data. The respective portions of the architecture may comprise various models that determine intermediate outputs that may be projected into a space associated with estimated cost. That estimated cost may identify an estimate of an output of the cost function for paths that are based on a particular prediction node of the tree search.