Object Motion Prediction Using Future Range Sensor Readings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional motion prediction techniques for autonomous driving are inaccurate due to noisy measurement data, which affects the reliability of predicting object motion and generating safe trajectories, especially in dynamic environments.
Innovation Solution
A method for predicting object motion based on predicted future sensor readings from range sensors, using a combination of historical data, point clustering, and multiple prediction methods, including Gaussian Process Approximation and Kalman filtering, to enhance accuracy and robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional techniques directly estimate motion based on current and historical motion data, then the prediction process is simple and fast, but the prediction accuracy deteriorates due to noisy measurement data
Solution Approach 1:
The patent applies preliminary action by computing predicted future sensor readings before performing motion estimation. The system predicts what sensor readings will be in the future based on current measurements, then uses these predicted readings for motion estimation. This preliminary computation of future sensor states resolves the contradiction by preparing refined data in advance, improving prediction accuracy while maintaining a structured processing pipeline.
Solution Approach 2:
The patent introduces an intermediary element - predicted future sensor readings - that mediates between raw noisy measurements and motion estimation. Instead of directly estimating motion from noisy current measurements, the system first generates predicted future sensor readings that serve as an intermediate representation. This intermediary data structure filters and refines the information, allowing accurate motion prediction without requiring overly complex processing of raw noisy data.
2Reliability
If conventional techniques use only current motion data for prediction, then the computation is fast and simple, but the reliability of prediction deteriorates in dynamic environments
Solution Approach 1:
The system performs preliminary computation by predicting future sensor readings ahead of time based on current measurements and historical data. This advance computation prepares refined prediction data that can be used reliably for motion estimation, improving prediction reliability in dynamic environments while maintaining efficient processing through pre-computation.
Solution Approach 2:
The patent implements feedback by using historical measurement data and current sensor readings to continuously refine predictions of future sensor states. The system feeds back past and present information into the prediction model, allowing it to adapt to dynamic environmental changes. This feedback mechanism improves prediction reliability by incorporating temporal patterns while managing computation time through efficient recursive updates.
3Measurement precision
If the system uses predicted future sensor readings for motion prediction, then the prediction accuracy improves, but the computational complexity increases
Solution Approach 1:
The system computes predicted future sensor readings as a preliminary step before motion estimation. By preparing these predicted readings in advance based on current measurements and historical data, the system improves motion prediction accuracy while organizing the computational workload into structured phases, managing overall computational complexity.
Solution Approach 2:
The predicted future sensor readings serve as an intermediary data structure that bridges raw sensor measurements and motion estimation. This intermediary representation consolidates complex computational operations into a standardized format, improving motion prediction accuracy while making the computational process more manageable and structured.
Data Source
AI summary
A method for predicting a motion of an object includes predicting the motion of the object based on predicted future sensor readings. The predicted future sensor readings are computed based on current measurement data of one or more range sensors.