ML Motion Planning for Vehicles Under Navigation Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional algorithms for route selection in vehicle navigation are impractical for complex routes and can lead to route overload when many vehicles use the same greedy algorithm, and factors like parked vehicles, construction zones, and pedestrians complicate navigation.
Innovation Solution
The use of motion planning with machine learning, where sensors on a vehicle extract feature vectors from environmental data and constraints to generate motion segments that minimize violations of these constraints, allowing the vehicle to operate efficiently and avoid obstacles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional greedy algorithms are used for route selection, then the route selection process is simple and fast, but the selected route becomes overloaded when many vehicles use the same algorithm and travel slows to a crawl
Solution Approach 1:
The patent implements feedback by having vehicles report their selected routes and travel times back to a central server. The server uses this feedback information to identify overloaded routes and generate alternative routes for future vehicle assignments, thereby preventing route overload while maintaining selection efficiency.
Solution Approach 2:
The patent introduces a central server as an intermediary between vehicles and route selection. Instead of each vehicle independently selecting routes using greedy algorithms, the server acts as a mediator that coordinates route assignments across the fleet, distributing traffic more evenly and preventing overload on any single route.
2Reliability
If complex routes with many decisions are selected to meet objectives like maximum driving time, then navigation objectives can be met, but traditional algorithms become impractical
Solution Approach 1:
The patent replaces traditional mechanical algorithmic approaches with machine learning models. Instead of using complex decision-making algorithms to evaluate multiple route options, the patent trains machine learning circuits on historical navigation data to automatically predict optimal routes that meet objectives like maximum driving time, significantly reducing computational complexity.
Solution Approach 2:
The patent performs preliminary action by pre-training machine learning models on extensive historical navigation data before actual route selection. This pre-processing allows the system to quickly predict optimal routes during operation without needing to perform complex real-time calculations, making complex route planning practical.
3Reliability
If the vehicle operates with motion planning that minimizes constraint violations, then safety and passenger comfort are enhanced, but the number of possible motion segments increases
Solution Approach 1:
The patent applies segmentation by dividing the continuous motion planning problem into discrete motion segments. Each motion segment represents a standardized maneuver (e.g., accelerate, decelerate, maintain speed) that can be independently evaluated against constraints. This segmentation allows the machine learning model to select from predefined segments rather than optimizing continuous parameters, reducing complexity while maintaining safety.
Solution Approach 2:
The patent changes parameters by transforming the motion planning problem from continuous parameter optimization to discrete parameter selection. Instead of optimizing continuous speed and acceleration values, the system selects from discrete motion segments with predefined parameters, simplifying the planning process while still achieving safe and comfortable operation through constraint-based filtering.
Data Source
AI summary
Techniques for operation of a vehicle using machine learning with motion planning include storing, using one or more processors of a vehicle located within an environment, a plurality of constraints for operating the vehicle within the environment. One or more sensors of the vehicle receive sensor data describing the environment. The one or more processors extract a feature vector from the stored plurality of constraints and the received sensor data. The feature vector includes a first feature describing an object located within the environment. A machine learning circuit of the vehicle is used to generate a first motion segment based on the feature vector. A number of violations of the stored plurality of constraints is below a threshold. The one or more processors operate the vehicle in accordance with the generated first motion segment.