Autonomous Vehicle Navigation Using Stochastic Games in Crowded Traffic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving systems in crowded environments fail to effectively account for actions and probabilities of opposing vehicles, limiting their ability to adapt and navigate safely and efficiently.
Innovation Solution
A computer-implemented method and system that utilize a stochastic game framework to determine an action space for ego and target vehicles, training a neural network with reward data to control vehicle navigation and avoid collisions in crowded environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional sensor-based autonomous driving systems are used, then real-time object detection is achieved, but the system cannot predict actions of opposing vehicles in crowded environments
Solution Approach 1:
The system performs preliminary action by training a neural network model in advance using stochastic games that simulate various driving scenarios. The model learns to predict opposing vehicle actions and determine safe navigation strategies before actual deployment, enabling it to anticipate and respond to potential conflicts in crowded environments without real-time computational burden
Solution Approach 2:
The system creates a virtual copy of the crowded environment through stochastic game simulations, where virtual vehicles replicate the behavior patterns of real opposing vehicles. This copying approach allows the neural network to learn from simulated interactions without requiring actual real-time data from opposing vehicles, thereby predicting their actions based on learned patterns
2Adaptability or versatility
If stochastic game framework with neural network training is implemented, then prediction of opposing vehicle actions is improved, but computational complexity increases
Solution Approach 1:
The complex neural network training and stochastic game execution are performed as preliminary actions during an offline training phase. Once trained, the model contains pre-learned knowledge that can be deployed with reduced computational complexity during real-time autonomous driving operations, separating the heavy computational burden from the actual navigation task
Solution Approach 2:
The system extracts the complex computational processes (stochastic game execution and neural network training) as a separate preprocessing stage. This extraction allows the main autonomous driving system to use the pre-computed predictions and policies without directly implementing the complex training framework, thereby reducing overall system complexity while maintaining adaptability
3Reliability
If vehicle navigation adapts to opposing vehicles in real-time, then safety in crowded environments is improved, but navigation speed may be reduced
Solution Approach 1:
The neural network is trained in advance to pre-compute navigation policies for various crowded environment scenarios. During actual navigation, the system quickly queries the pre-trained model for action predictions rather than performing complex real-time calculations, thereby maintaining both safety through adaptive prediction and speed through efficient inference
Solution Approach 2:
The system performs partial action by focusing computational resources on predicting only the critical opposing vehicle actions that impact navigation safety, rather than analyzing all possible environmental factors. This selective approach maintains navigation speed while achieving sufficient safety through targeted predictions of the most relevant vehicle behaviors
Data Source
AI summary
A system and method for providing autonomous vehicular navigation within a crowded environment that include receiving data associated with an environment in which an ego vehicle and a target vehicle are traveling. The system and method also include determining an action space based on the data associated with the environment. The system and method additionally include executing a stochastic game associated with navigation of the ego vehicle and the target vehicle within the action space. The system and method further include controlling at least one of the ego vehicle and the target vehicle to navigate in the crowded environment based on execution of the stochastic game.