Neural Actuation Mapping for Adaptive Vehicle Motion Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous vehicle control systems rely on static actuation maps, which are limited by interpolation between predefined set-points and require high-precision sensors, making them less adaptable and more dependent on specific environmental conditions.
Innovation Solution
Implementing a dynamic actuation map using a neural network fed by visual odometry, which learns the actuation map during training, provides a continuous and precise map that adapts in real-time to changing vehicle dynamics without the need for high-precision sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a static actuation map with look-up table is used, then the control system is simple to implement, but the adaptability to changing vehicle dynamics is poor
Solution Approach 1:
The patent implements a dynamic actuation map using a neural network that continuously adapts to changing vehicle dynamics during operation. The neural network model is trained offline with vehicle dynamics data and then deployed to generate actuation commands that automatically adjust to varying conditions, transforming a static control system into a dynamic one that evolves with the vehicle's state.
Solution Approach 2:
The patent changes the parameters of the actuation map from fixed look-up table values to continuous neural network predictions. The neural network takes vehicle state parameters (position, velocity, acceleration) and dynamically computes actuation commands, allowing the system to adapt to parameter changes in vehicle dynamics without requiring reconfiguration.
2Manufacturing precision
If interpolation is used in static actuation maps, then the control precision between set-points is improved, but the computational complexity and response time increase
Solution Approach 1:
The patent replaces the mechanical interpolation process with a neural network-based computational system. Instead of performing mathematical interpolation between discrete look-up table entries, the neural network directly predicts continuous actuation commands from vehicle state inputs, eliminating the need for interpolation while maintaining precision and reducing computational overhead.
3Measurement precision
If high-precision sensors are used, then the measurement accuracy of vehicle dynamics is improved, but the system cost and complexity increase
Solution Approach 1:
The patent replaces expensive high-precision sensors with standard, lower-cost sensors combined with a neural network processing system. The neural network compensates for the lower precision of standard sensors by learning from training data and providing robust actuation commands, effectively substituting hardware precision with software intelligence.
Solution Approach 2:
The neural network acts as an intermediary between standard sensors and the actuation system. It processes and interprets sensor data, compensating for measurement limitations and translating imperfect sensor readings into accurate actuation commands, thereby eliminating the need for high-precision sensors.
4Adaptability or versatility
If a dynamic actuation map with neural network is implemented, then the adaptability and precision are improved, but the computational resources required increase
Solution Approach 1:
The patent performs the computationally intensive neural network training offline before deployment. The trained model is then deployed to embedded systems where it requires minimal computational resources for real-time inference. This preliminary action transfers the heavy computational burden from the running system to the training phase, enabling real-time adaptation with low energy consumption.
Data Source
AI summary
A method of controlling actuators of a vehicle includes receiving pixel data corresponding to an area outside of the vehicle, receiving desired dynamic set-points for the vehicle from a path planning application of the vehicle, receiving sensor data from the vehicle, generating one or more real actuator commands for the vehicle in response to the sensor data from the vehicle, the desired dynamic set-points for the vehicle received from the path planning application of the vehicle, and one or more inferences made by a first trained neural network model using the sensor data from the vehicle and the desired dynamic set-points for the vehicle as input, and communicating the one or more real actuator commands for the vehicle to actuators of the vehicle. The one or more real actuator commands for the vehicle are generally presented at an output of the first trained neural network model. The first trained neural network model was generally trained by performing computer vision operations on the pixel data corresponding to the area outside the vehicle arranged as video frames to detect features in a first set of the video frames, determining a first set of measured dynamic set-points for the vehicle by applying visual odometry operations on the features detected in the first set of the video frames, and modifying a plurality of weights and bias values of an untrained neural network model based on the first set of measured dynamic set-points, the sensor data, and a training dataset comprising dynamic set-points and corresponding actuator values that are representative of one or more desired inferences.