Neural Network Energy Estimation for Autonomous Vehicle Trajectory Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous and semi-autonomous vehicles face delays and potential loss of functionality due to unpredictable battery performance affected by environmental conditions, such as temperature and wind, which can lead to incomplete voyages and safety issues.
Innovation Solution
The use of machine learning models, specifically recurrent neural networks (RNNs) and feed forward neural networks, to estimate battery state of charge and remaining operating time by training on historical data from various voyages, allowing for real-time adjustments in vehicle trajectory and refueling decisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If preemptive refueling scheduling is used, then vehicle operational continuity is improved, but prediction accuracy deteriorates due to unaccounted environmental conditions
Solution Approach 1:
The system performs preliminary refueling scheduling based on initial battery state predictions, then updates predictions during operation with actual environmental data. This allows advance planning while maintaining accuracy through continuous refinement as conditions are observed.
Solution Approach 2:
The system incorporates feedback loops where actual battery performance data and environmental conditions measured during voyages are used to refine and update prediction models. This feedback mechanism improves prediction accuracy for future refueling schedules while maintaining operational continuity.
2Ease of operation
If battery charge capacity is estimated without environmental factors, then calculation simplicity is improved, but prediction reliability deteriorates due to temperature and wind effects
Solution Approach 1:
The prediction system is segmented into multiple independent components: a base charge capacity calculator that provides simple estimates, and separate environmental adjustment modules for temperature, wind, and other factors. This allows the system to maintain calculation simplicity while incorporating reliability-improving environmental corrections when needed.
Solution Approach 2:
The system dynamically adjusts the complexity of charge capacity calculations based on operational context. During critical phases or when environmental data is available, the system activates environmental factor adjustments. During routine operations or when data is unavailable, it uses simpler calculations to maintain ease of operation.
3Reliability
If environmental conditions are monitored and incorporated into predictions, then voyage completion reliability is improved, but system complexity increases
Solution Approach 1:
The system uses a universal prediction framework that can handle multiple environmental factors (temperature, wind, humidity, pressure) through a single integrated model. This multi-functional approach improves voyage completion reliability across diverse conditions without proportionally increasing system complexity, as the same core infrastructure processes all environmental inputs.
Solution Approach 2:
The prediction system automatically monitors environmental conditions, processes the data, and adjusts charge capacity estimates without requiring manual intervention. The system self-calibrates using historical data and environmental patterns, reducing the operational complexity burden despite the enhanced predictive capabilities.
4Measurement precision
If real-time battery state monitoring is implemented, then energy management precision is improved, but computational load increases
Solution Approach 1:
The system implements partial real-time monitoring by continuously tracking only the most critical battery parameters (charge level, temperature) while periodically sampling less critical metrics. This selective monitoring approach maintains sufficient energy management precision while reducing computational load and power consumption compared to comprehensive continuous monitoring of all battery characteristics.
Data Source
AI summary
Controlling a vehicle according to a trained neural network model capable of being used to generate an output from which one or more vehicle operating variables can be estimated. The neural network model can be used to process, as input, aggregated data corresponding to operational and/or environmental characteristics experienced by the vehicle during at least a portion of a voyage. The aggregated data can include a range of values collected over a period of time when the vehicle is traversing the portion of the voyage. The output generated by the neural network model, based on processing the input, can be further processed in order to determine, for example, an estimated state of charge and/or an estimated remaining flight time for the vehicle. Such estimated values can thereafter be used by a controller of the vehicle to maintain course or maneuver to a charging station.