Autonomous Vehicle AI Navigation Using Holistic Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous vehicle navigation systems face high failure rates due to independent sub-systems that do not consider holistic goals and make inaccurate predictions, especially in complex environments with unforeseen changes.
Innovation Solution
An AI system that generates an action-value array using reinforcement learning to guide the vehicle towards a target by synchronizing sensor data, transforming it into a common format, and applying a navigation policy to estimate desirable navigational actions, while avoiding obstacles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If separate sub-systems are used for object detection, tracking, and free space detection, then each sub-system can be developed independently with defined tasks, but the system reliability decreases and failure rate increases when subsystems fail or make incorrect assumptions
Solution Approach 1:
The patent merges multiple independent sub-systems (object detection, tracking, free space detection) into a unified deep learning model that processes sensor data holistically. This integration eliminates the need for separate subsystems to make independent assumptions about distance and characteristics, thereby improving reliability while maintaining the ability to perform specialized functions through integrated feature extraction and decision-making layers.
2Adaptability or versatility
If machine learning models predict actuation commands directly from sensor information, then the system can handle complex environments, but the prediction accuracy decreases and the system becomes deterministic unable to adapt to sudden changes
Solution Approach 1:
The patent implements a dynamic navigation system using deep reinforcement learning that continuously adapts to changing environments. The model processes raw sensor data through multiple convolutional layers to extract features dynamically, and the reinforcement learning component learns optimal policies that adapt to sudden changes in road conditions, weather, and obstacles, thereby improving both prediction accuracy and adaptability to complex environments.
Data Source
AI summary
An artificial intelligence system for an autonomous vehicle includes one or more processors and a memory in communication with the one or more processors and storing a target-orientated navigation system module. When executed, the target-orientated navigation system module causes the one or more processors to receive sensor data from one or more sensors of a vehicle, synchronize the sensor data, preprocess the synchronized sensor data by transforming the sensor data into a common data format, concatenate the transformed sensor data into a K-dimensional array, which acts as an input state array, apply a navigation policy to the input state array to estimate an action-value array, direct a vehicle control system to guide the vehicle to a location representative of a cell in the action-value array that has a highest reward value.