Probabilistic Lane Planning for Uncertain Autonomous Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional lane planning in autonomous vehicles operates with a deterministic approach, limiting the set of actions available and failing to account for uncertainties such as road or traffic conditions, leading to inefficient route planning and potential safety issues.
Innovation Solution
A lane planner that generates output data based on a probabilistic action space, using a hierarchical drive planning framework and reinforcement learning to calculate expected time rewards, allowing for more flexible and adaptive route planning by considering the probability of action success and failure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a deterministic approach is used for lane planning with a limited set of actions, then the planning process is simple and computationally efficient, but the system cannot adapt to uncertain road or traffic conditions and may fail to execute required actions
Solution Approach 1:
The patent applies dynamics by transitioning from a static deterministic action model to a dynamic probabilistic action model. The lane planner now considers multiple possible actions with associated probabilities of success, allowing the system to adapt to uncertain conditions. The action space includes probabilistic outcomes where actions may succeed or fail based on traffic conditions, requiring dynamic replanning and selection of alternative actions.
Solution Approach 2:
The patent changes the fundamental parameter of action representation from deterministic (single outcome) to probabilistic (multiple outcomes with probabilities). This parameter change enables the system to model uncertainty in action execution, where each action has a probability of success rather than guaranteed execution, allowing adaptation to varying traffic and road conditions.
2Reliability
If a deterministic graph search algorithm is used to find the shortest path, then the computation is efficient and straightforward, but the system does not account for action failures or blocking by other vehicles
Solution Approach 1:
The patent applies preliminary action by pre-computing not just the shortest path but multiple alternative paths with their associated probabilities. The system prepares a set of candidate actions and paths in advance, considering potential failures, so that when an action is blocked or fails, alternative pre-planned paths can be executed without requiring complete replanning from scratch.
Solution Approach 2:
The patent implements feedback mechanisms where the actual execution outcomes of actions are fed back into the planning system. When an action fails or is blocked, this feedback triggers updates to the probability distributions and may initiate replanning to select alternative actions, creating a closed-loop system that continuously adapts based on execution results.
3Adaptability or versatility
If the lane planner provides a limited set of deterministic actions, then the output is simple and easy to execute, but the vehicle may get stuck when actions are blocked by traffic conditions
Solution Approach 1:
The patent makes the action selection dynamic by maintaining a distribution over multiple possible actions rather than selecting a single deterministic action. The system can dynamically shift probability mass between different actions based on current traffic conditions and observed failures, allowing flexible adaptation without requiring complete replanning when conditions change.
Solution Approach 2:
The patent applies partial action by providing a set of candidate actions with varying probabilities rather than a single full-commitment action. This allows the system to attempt actions with lower confidence while having backup options ready, effectively preparing multiple partial plans that can be executed in sequence or combination depending on outcomes.
Data Source
AI summary
In various examples, a lane planner for generating lane planner output data based on a state and probabilistic action space is provided. A driving system—that operates based on a hierarchical drive planning framework—includes the lane planner and other planning and control components. The lane planner processes lane planner input data (e.g., large lane graph, source node, target node) to generate lane planner output data (e.g., expected time rewards). The driving system can also include a route planner (e.g., a first planning layer) that operates to provide the lane planner input data to the lane planner. The lane planner operates as second planning layer that processes the lane planner input data based at least in part on a state and probabilistic action space of the large lane graph and calculates a time cost associated with navigating from a source node to a target node in the large lane graph.