Autonomous Vehicle AI Model Retraining via Passenger Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in adapting their control logic to improve navigation and collision avoidance based on real-time passenger feedback, as existing systems rely heavily on pre-programmed algorithms that do not account for subjective experiences or anomalies during rides.
Innovation Solution
Implementing an AI model, such as a deep reinforcement learning algorithm, that updates control logic by integrating passenger feedback and sensor data to refine obstacle avoidance strategies, allowing the system to learn from ride experiences and adjust vehicle behavior to enhance passenger satisfaction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If pre-programmed algorithms are used for autonomous vehicle control, then the system structure is simple and reliable, but the system cannot adapt to subjective passenger experiences or ride anomalies
Solution Approach 1:
The patent implements a feedback mechanism where passenger ratings and ride anomaly reports are collected and used to retrain the neural network model. The controller receives feedback data from passengers about their ride experiences, processes this feedback, and uses it to update the control logic through neural network retraining, enabling continuous adaptation to passenger preferences and improving the system's adaptability over time.
Solution Approach 2:
The system performs self-learning and self-improvement by automatically collecting ride data, processing passenger feedback, and retraining its own neural network model without requiring external manual programming. The autonomous vehicle controller continuously updates its control logic based on accumulated ride experiences, enabling the system to serve itself in terms of adaptation and improvement.
2Reliability
If pre-programmed control logic is used, then the system is reliable and stable, but it cannot learn from ride experiences or improve navigation strategies
Solution Approach 1:
The system is pre-equipped with a neural network model and the capability to collect and process feedback data during rides. The controller is designed in advance to receive passenger ratings and anomaly reports, store this feedback data, and systematically retrain the neural network using this accumulated information, enabling continuous improvement while maintaining operational reliability.
Solution Approach 2:
The control logic transitions from a static pre-programmed system to a dynamic system that evolves over time. The neural network model is continuously updated through retraining with new feedback data, allowing the control logic to adapt and improve its navigation and collision avoidance strategies based on real-world ride experiences while maintaining stable operation.
3Adaptability or versatility
If the system collects and processes passenger feedback in real-time, then the adaptability improves, but the processing time and computational load increase
Solution Approach 1:
The system collects feedback data continuously during rides but processes and retrains the neural network periodically rather than in real-time for each input. The controller accumulates feedback data from multiple rides and performs batch retraining operations, balancing adaptability with acceptable processing time by updating the model at regular intervals rather than continuously.
Data Source
AI summary
A controller receives sensor data during a ride and provides it to a server system. A passenger further provides feedback concerning the ride in the form of some or all of an overall rating, flagging of ride anomalies, and flagging of road anomalies. The sensor data and feedback are input to a training algorithm, such as a deep reinforcement learning algorithm, which updates an artificial intelligence (AI) model. The updated model is then propagated to controllers of one or more autonomous vehicle which then perform autonomous navigation and collision avoidance using the updated AI model.