Autonomous Vehicle Trajectory Planning with Confidence-Based Obstacle Shapes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving vehicles face challenges in planning trajectories to avoid obstacles when the prediction of the obstacle's future position is not confident, leading to potential collisions.
Innovation Solution
The method involves detecting obstacles using sensor data, predicting a distribution of possible positions based on confidence levels, determining a modified shape with a modified length of the obstacle, and planning a trajectory accordingly to ensure safe navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the ADV plans trajectory based on a single predicted position of the obstacle, then the planning process is simple and fast, but the collision risk increases due to prediction uncertainty
Solution Approach 1:
The system performs preliminary action by expanding the obstacle's predicted position into a modified shape that encompasses multiple possible future positions before trajectory planning begins. This pre-computation of position uncertainty allows the planner to treat the expanded shape as a single planning target, avoiding the need to evaluate multiple separate trajectories while maintaining high collision avoidance reliability.
Solution Approach 2:
The patent applies dimensionality change by transforming the obstacle from a single-point prediction in space to a modified shape that includes spatial uncertainty dimensions. The modified shape extends beyond the predicted center position to include boundary points representing possible deviation ranges, effectively adding a dimension of uncertainty representation to the planning problem.
2Reliability
If the ADV uses a modified shape with extended length to account for prediction uncertainty, then collision avoidance improves, but the required safety distance increases
Solution Approach 1:
The system applies local quality by differentiating the obstacle representation into different regions: the center point represents the most likely position, while the extended boundaries represent uncertainty ranges. The trajectory planner can focus on avoiding the center while the modified shape automatically accounts for local uncertainty variations, allowing safer navigation without uniformly increasing distance from all points.
Solution Approach 2:
The patent changes parameters by transforming the obstacle from a static single-position object to a dynamic modified shape with extended length that reflects prediction confidence. The parameter change allows the system to encode uncertainty magnitude directly into the geometric representation, where the extension length correlates with prediction uncertainty rather than requiring separate uncertainty parameters.
3Reliability
If the ADV plans trajectory avoiding the entire modified shape, then safety is maximized, but the vehicle's maneuverability and speed are reduced
Solution Approach 1:
The system performs preliminary action by pre-computing the modified shape that encompasses all plausible obstacle positions before trajectory optimization. This allows the trajectory planner to work with a fixed uncertainty boundary and optimize speed and path efficiently, rather than conservatively slowing down to account for uncertainty during real-time planning.
Solution Approach 2:
The patent applies dynamics by making the modified shape's extension dynamic rather than static. The shape adjusts based on prediction confidence and uncertainty estimates, allowing the vehicle to maintain higher speeds when uncertainty is low (smaller modified shape) and increase caution only when necessary (larger modified shape), optimizing the trade-off between safety and speed.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
An obstacle is detected based on sensor data obtained from a plurality of sensors of the ADV. A distribution of a plurality of positions of the obstacle at a point of time may be predicted. A range of positions of the plurality of positions of the obstacle may be determined based on a confidence level of the range. A modified shape with a modified length of the obstacle may be determined based on the range of positions of the obstacle. A trajectory of the ADV based on the modified shape with the modified length of the obstacle may be planned. The ADV may be controlled to drive according to the planned trajectory to drive safely to avoid a collision with the obstacle.