Waypoint-Guided Trajectory Prediction for Smooth Lane Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing trajectory prediction methods in autonomous driving face issues such as insufficient lateral avoidance and deceleration, particularly in scenarios like lane changes and overtaking, due to multi-modal trajectories or discontinuous predictions across sub-segments, leading to unpredictable vehicle behavior.
Innovation Solution
A method that involves obtaining multiple waypoint sequences on a future route segment, performing trajectory prediction under the guidance of these sequences, and determining a target trajectory based on scoring, which enhances the reasonableness of predictions by avoiding discontinuities and ensuring adequate lateral maneuvers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple waypoint sequences are used for trajectory prediction, then the accuracy and consistency of predictions improve, but the device complexity increases
Solution Approach 1:
The trajectory prediction problem is segmented into multiple waypoint sequences, where each sequence represents a different possible path. The system divides the prediction task into generating multiple candidate trajectories and then selecting the optimal one, rather than attempting to predict a single trajectory. This segmentation allows the system to explore multiple possibilities and improve prediction accuracy while managing complexity through structured processing of each candidate path.
Solution Approach 2:
The patent introduces an intermediary scoring mechanism that evaluates multiple predicted trajectories and selects the optimal one. This intermediary layer (the scoring and selection module) acts as a mediator between the multiple candidate trajectories generated by the prediction model and the final selected trajectory. The scoring function provides a systematic way to compare and rank different trajectory options, improving consistency without requiring the prediction model itself to become overly complex.
2Reliability
If trajectory prediction is performed under guidance of multiple waypoint sequences, then lateral avoidance and deceleration improve, but the calculation time increases
Solution Approach 1:
The system performs preliminary action by pre-generating multiple waypoint sequences that represent potential future paths. These waypoint sequences are prepared in advance as candidate trajectories, allowing the prediction model to work with pre-structured data rather than generating all possibilities from scratch during real-time prediction. This preliminary structuring of possible paths enables better lateral avoidance capabilities while reducing the computational burden during the actual prediction phase.
Solution Approach 2:
The patent applies partial action by generating multiple waypoint sequences and candidate trajectories, but then using a scoring mechanism to select only the most promising candidates for final selection. Rather than exhaustively evaluating every possible trajectory combination, the system performs partial evaluation on multiple candidates and selects the top-scoring ones. This approach provides sufficient lateral avoidance capability through multiple candidates while avoiding the excessive calculation time that would result from evaluating all possible trajectories in detail.
3Stability of the object's composition
If scores are assigned to multiple predicted trajectories for selection, then the consistency of navigation improves, but the processing complexity increases
Solution Approach 1:
The patent uses parameter changes by introducing a scoring parameter that quantifies the quality of each predicted trajectory. This scoring parameter transforms the complex multi-dimensional comparison of different trajectories into a single comparable metric. By changing the evaluation from a complex qualitative assessment to a quantitative scoring system, the patent achieves consistent navigation selection while managing processing complexity through standardized parameter-based comparison.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure provides a trajectory prediction method, a method for training a trajectory prediction model, a medium, and a device, and relates to the technical field of computers, and in particular to the technical fields of autonomous driving and artificial intelligence. An implementation solution includes: obtaining first information and a plurality of waypoint sequences of a target vehicle, wherein the first information comprises navigation information of the target vehicle and perception information of surroundings of the target vehicle, and wherein each waypoint sequence of the plurality of waypoint sequences comprises a plurality of waypoints within a future route segment; performing, for each waypoint sequence of the plurality of waypoint sequences, trajectory prediction based on the first information under the guidance of the waypoint sequence to obtain a predicted trajectory corresponding to the waypoint sequence and a score for the predicted trajectory; and determining, based on respective scores for a plurality of predicted trajectories corresponding to the plurality of waypoint sequences, a target predicted trajectory among the plurality of predicted trajectories.