Neural Route Segment Selection for ODD-Aware Vehicle Expeditions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle expedition route planning is inefficient due to lack of awareness of near-future traffic situations and limited knowledge of traffic environments, leading to suboptimal data collection within Operational Design Domain (ODD) requirements.
Innovation Solution
A route planning system utilizing a neural network trained on real-time traffic information and vehicle geolocation to select road segments that maximize travel within ODD requirements, optimizing route planning through machine learning and dynamic traffic assessment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional route planning methods are used by operators or drivers, then route planning can be performed in advance, but the efficiency of data collection is reduced due to lack of real-time traffic information and limited knowledge of traffic environments
Solution Approach 1:
The neural network is trained in advance using historical traffic data and digital map information to learn optimal routing patterns. This preliminary training enables the system to make rapid real-time decisions without requiring real-time human intervention, thus improving data collection efficiency while incorporating learned knowledge about traffic environments
Solution Approach 2:
The system continuously receives real-time traffic information from sensors and external sources, feeds this information back to the neural network, and dynamically adjusts the route selection. This feedback loop ensures that the routing decisions are based on current traffic conditions rather than static pre-planned routes, eliminating the loss of real-time traffic information
2Adaptability or versatility
If traditional route planning is used with pre-determined routes, then planning can be done in advance, but the ability to adapt to current traffic situations is poor
Solution Approach 1:
The patent replaces traditional mechanical route planning methods (manual operator planning or simple GPS navigation) with an intelligent neural network system. This substitution enables the system to process complex real-time traffic information and adapt routing decisions dynamically, significantly improving adaptability to traffic situations. The neural network, once trained, provides adaptive capabilities without requiring complex real-time computation infrastructure
Solution Approach 2:
The system changes the parameter representation of route planning by using neural network weights and activation states instead of traditional routing algorithms. By adjusting neural network parameters based on real-time traffic inputs, the system achieves high adaptability. The trained neural network model transforms static routing parameters into dynamic, adaptive routing decisions that respond to changing traffic conditions
3Productivity
If more computing resources and storage capacity are allocated for route planning, then better route optimization is possible, but the carbon footprint increases
Solution Approach 1:
The computationally intensive neural network training is performed in advance during off-peak hours or when energy availability is abundant. This preliminary action transfers the energy consumption from the operational phase to the training phase, allowing the deployed system to make routing decisions with minimal real-time computational resources, thus reducing the carbon footprint during actual data collection expeditions while maintaining high route optimization quality
Solution Approach 2:
The system uses a trained neural network model that provides sufficient routing optimization without requiring continuous full-power computation. By deploying a pre-trained model that makes predictions with limited real-time computational resources, the system achieves adequate route optimization quality (partial action sufficient for needs) rather than continuously maximizing optimization, thereby reducing energy consumption and carbon footprint during operation
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure relates to a method performed by a route planning system (1) of a vehicle (2) for road segment selection along a route to be travelled by the vehicle. The route planning system determines (1001) with support from a positioning system (22), a geolocation of the vehicle in view of a digital map (3). The route planning system further identifies (1002) in the digital map based on the vehicle geolocation, an upcoming road junction (51) which the vehicle is approaching and/or is located at, which upcoming road junction comprises two or more upcoming road segments (41). Moreover, the route planning system derives (1003) traffic information data (6) applicable for a map area of the digital map covering a plurality of road segments (4) including the two or more upcoming road segments. Furthermore, the route planning system selects (1004) a road segment (410) out of the two or more upcoming road segments by feeding one or more parameters related to the traffic information data and one or more parameters related to the vehicle geolocation through a neural network trained to select the road segment rendering greatest extent of travelling within one or more set Operational Design Domain, ODD, requirements. The disclosure also relates to a route planning system in accordance with the foregoing, a vehicle comprising such a route planning system, and a respective corresponding computer program product and non-volatile computer readable storage medium.