Autonomous Vehicle Speed Optimization via Reinforcement Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles programmed to simply navigate from an origin to a destination may not optimize speed, potentially leading to delayed arrival times due to traveling with slower-moving vehicles, as they do not maneuver to optimize speed.
Innovation Solution
A reinforcement learning reward function is developed to optimize vehicle speed by identifying and maneuvering to target areas based on relative velocity and length, incorporating safety penalties and considering multiple lanes, with a computer system using sensors to select target areas for efficient navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If the autonomous vehicle is programmed to simply navigate from an origin to a destination, then the navigation is straightforward and simple, but the vehicle arrives later than optimal due to not optimizing speed
Solution Approach 1:
The patent applies dynamics by transitioning from static, pre-programmed navigation routes to dynamic, real-time route optimization. The system continuously adjusts the vehicle's path based on current traffic conditions, identifying optimal target areas ahead of the vehicle and dynamically recalculating routes to maximize speed while navigating through traffic lanes.
Solution Approach 2:
The system implements feedback by continuously monitoring the vehicle's position, speed, and surrounding traffic conditions, then using this information to identify optimal target areas and adjust the navigation path in real-time. This closed-loop control enables the vehicle to adapt to changing traffic conditions and optimize arrival time dynamically.
2Loss of time
If the vehicle maneuvers to optimize speed by selecting target areas in multiple lanes, then the travel time is reduced, but the computational overhead increases
Solution Approach 1:
The patent applies segmentation by dividing the road ahead into discrete target areas rather than calculating continuous optimal paths. The system identifies specific zones ahead of the vehicle where lane changes or maneuvers would be beneficial, breaking down the complex navigation problem into manageable discrete decisions about which target areas to select.
Solution Approach 2:
The system uses partial action by focusing computational resources on identifying only the most critical target areas ahead of the vehicle rather than evaluating all possible future positions. This selective approach reduces computational overhead while still achieving significant speed optimization by concentrating on key decision points in the navigation path.
3Speed
If the vehicle aggressively maneuvers to reach target areas faster, then the speed is optimized, but the safety risk increases
Solution Approach 1:
The patent applies preliminary action by identifying and planning for target areas ahead of the vehicle's current position, allowing the system to prepare optimal maneuvers in advance. By calculating desired trajectories and selecting target areas before the vehicle reaches critical decision points, the system can execute smoother, safer maneuvers rather than reacting abruptly to immediate conditions.
Solution Approach 2:
The system implements beforehand cushioning by incorporating safety constraints and penalty functions into the target area selection process. The reward function includes safety considerations that penalize aggressive or risky maneuvers, effectively cushioning against safety risks before they occur by designing the decision-making framework to inherently favor safer options.
Data Source
AI summary
One or more target areas are identified proximate to a moving vehicle. The vehicle can be maneuvered to a target area selected according to a reinforcement learning reward function.