Autonomous Vehicle Sideslip Prediction and Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately controlling turns due to sideslip or skid, which can lead to control errors and safety issues, especially on varying road conditions, and lack sophisticated methods for predicting vehicle status based on current conditions.
Innovation Solution
A sideslip predictive model is developed using a physical model and machine-learning techniques to predict sideslip based on driving parameters and conditions, allowing for compensation in planning and control data to adjust vehicle movement and prevent skidding, with a scenario/sideslip mapping table created from recorded driving statistics to inform real-time adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a vehicle makes turns on varying road conditions, then the vehicle can navigate different locations, but sideslip or skid occurs causing control errors and safety issues
Solution Approach 1:
The system predicts sideslip angle in advance using a neural network model before the vehicle executes turning maneuvers. By forecasting the sideslip angle based on current driving parameters (steering wheel angle, yaw rate, lateral acceleration, wheel velocities), the control system can pre-adjust steering commands and speed to compensate for anticipated sideslip, preventing control errors before they occur
Solution Approach 2:
The system continuously monitors actual driving parameters (steering wheel angle, yaw rate, lateral acceleration, wheel velocities) and feeds this information back to the neural network prediction model. This closed-loop feedback enables the model to adapt to varying road conditions in real-time, continuously refining sideslip predictions and control adjustments to maintain accuracy across different navigation scenarios
2Adaptability or versatility
If different vehicles run on different roads with different speeds and steering angles, then vehicles can adapt to various driving scenarios, but turning radius varies impacting control accuracy
Solution Approach 1:
The neural network prediction model dynamically adjusts its predictions based on changing driving parameters including vehicle speed, steering wheel angle, yaw rate, lateral acceleration, and wheel velocities. By continuously monitoring and responding to parameter changes, the system maintains accurate sideslip predictions across diverse driving scenarios with different vehicles, roads, and speeds
Solution Approach 2:
The system performs preliminary prediction of sideslip angle using the neural network model before control execution, allowing the system to pre-calculate appropriate control adjustments for varying turning radii. This advance prediction enables the control system to compensate for differences in turning radius caused by various vehicle configurations and road conditions
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A set of driving scenarios are determined for different types of vehicles. Each driving scenario corresponds to a specific movement of a particular type of autonomous vehicles. For each of the driving scenarios of each type of autonomous vehicles, a set of driving statistics is obtained, including driving parameters used to control and drive the vehicle, a driving condition at the point in time, and a sideslip caused by the driving parameters and the driving condition under the driving scenario. A driving scenario/sideslip mapping table or database is constructed. The scenario/sideslip mapping table includes a number of mapping entries. Each mapping entry maps a particular driving scenario to a sideslip that is calculated based on the driving statistics. The scenario/sideslip mapping table is utilized subsequently to predict the sideslip under the similar driving environment, such that the driving planning and control can be compensated.