Vehicle Transmission Gear Control Using Reinforcement Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional transmission control systems require numerous shift patterns and extensive testing to prevent busy shifting, making it difficult to implement efficient gear stage control in vehicles.
Innovation Solution
A transmission control apparatus and method using reinforcement learning to determine an optimal gear stage based on driving, road, and driver information, with a controller assigning rewards to the agent for fuel consumption, accelerator, brake, and engine RPM to optimize gear shifts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional shift patterns are used to control transmission, then gear stage control can be implemented, but hundreds of patterns are required and extensive testing time is needed
Solution Approach 1:
The patent changes the control parameters from discrete shift patterns to continuous reinforcement learning parameters. The agent learns optimal gear stage decisions by receiving rewards based on fuel consumption, accelerator position, brake position, and engine RPM, replacing the need for hundreds of predefined shift patterns with a dynamic learning-based approach.
Solution Approach 2:
The transmission control system performs self-optimization through the reinforcement learning agent that automatically learns optimal shift patterns by interacting with the vehicle environment. The agent receives feedback in the form of rewards or penalties and autonomously improves its decision-making without requiring extensive external testing and manual pattern definition.
2Adaptability or versatility
If multiple shift patterns are provided to cover various driving conditions, then comprehensive gear control is achieved, but test and experimentation time increases significantly
Solution Approach 1:
The reinforcement learning agent receives continuous feedback in the form of rewards or penalties from the controller based on its gear stage decisions. The feedback is generated from real-time vehicle data including fuel consumption, accelerator position, brake position, and engine RPM, allowing the agent to learn and adapt to different driving conditions dynamically without extensive pre-testing.
Solution Approach 2:
The agent performs preliminary learning during normal vehicle operation by continuously receiving feedback and updating its policy. This ongoing learning process allows the system to adapt to various driving conditions in real-time, eliminating the need for extensive separate testing phases before deployment.
3Reliability
If separate upshift and downshift patterns are used to prevent busy shifting, then busy shift phenomenon is prevented, but the control system becomes more complex
Solution Approach 1:
The reinforcement learning agent serves as a universal decision-making mechanism that handles both upshift and downshift decisions through a single integrated policy. The agent evaluates the current state and determines the optimal gear stage regardless of whether an upshift or downshift is needed, replacing the need for separate upshift and downshift pattern logic.
Data Source
AI summary
An apparatus and a method for controlling a transmission of a vehicle include an agent that determines a gear stage of the vehicle based on driving information of the vehicle, shape information of a road, and/or operation information of a driver, and include a controller that performs a reward with respect to the determination of the agent and controls the transmission of the vehicle based on the gear stage determined by the agent.