Motion Planning Uncertainty Convolution for Autonomous Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately predicting the motion of objects in their environment, particularly in uncertain situations, which can lead to unsafe navigation and reduced efficiency in motion planning.
Innovation Solution
A computer-implemented method that utilizes a machine-learned system to convert object shapes into probabilities of occupancy by convolving occupied areas with continuous uncertainty, determining the probability of future occupancy of locations based on motion predictions and object detections, and providing this data to a motion planning system for comprehensive reasoning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional motion planning methods are used in autonomous vehicles, then the system complexity is reduced, but the accuracy of predicting object motion and the safety of navigation deteriorate due to inability to handle full uncertainty
Solution Approach 1:
The system segments the uncertainty representation into discrete probability distributions over continuous state spaces. Instead of handling continuous uncertainty directly, the method divides the state space into discrete bins and assigns probability masses to each bin, enabling tractable computation while preserving uncertainty information.
Solution Approach 2:
The method transforms the representation of uncertainty from continuous probability density functions to discrete probability distributions. This parameter change allows the system to reason about full uncertainty using discrete mathematics while maintaining computational efficiency through binning and probability mass assignment.
2Measurement precision
If continuous uncertainty is fully represented and reasoned about in motion planning, then the accuracy of motion predictions improves, but the computational complexity increases
Solution Approach 1:
The continuous state space is segmented into discrete bins, and probability masses are assigned to each bin. This segmentation allows the system to represent continuous uncertainty in a computationally tractable discrete form, enabling accurate motion predictions without overwhelming computational complexity.
Solution Approach 2:
The method substitutes complex continuous probability density function operations with discrete probability distribution manipulations. By replacing continuous mathematical operations with discrete computations over binned states, the system achieves accurate uncertainty reasoning with reduced computational burden.
3Productivity
If discrete probability distributions are used to represent continuous uncertainty, then the computational tractability improves, but the measurement precision of uncertainty representation may be reduced
Solution Approach 1:
The system applies local quality by using finer binning in regions of high probability mass and coarser binning in regions of low probability mass. This adaptive discretization preserves measurement precision where it matters most (in high-probability regions) while maintaining computational tractability overall.
Solution Approach 2:
The discretization scheme is made dynamic by adapting the bin boundaries and resolutions based on the underlying probability distribution. Regions with higher probability density receive finer discretization, while low-probability regions use coarser discretization, optimizing the trade-off between precision and computational efficiency.
Data Source
AI summary
Systems and methods for motion planning by a vehicle computing system of an autonomous vehicle are provided. The vehicle computing system can input sensor data to a machine-learned system including one or more machine-learned models. The computing system can obtain, as an output of the machine-learned model(s), motion prediction(s) associated with object(s) detected by the system. The system can convert a shape of the object(s) into a probability of occupancy by convolving an occupied area of the object(s) with a continuous uncertainty associated with the object(s). The system can determine a probability of future occupancy of a plurality of locations in the environment at future times based at least in part on the motion prediction(s) and the probability of occupancy of the object(s). The system can provide the motion prediction(s) and the probability of future occupancy of the plurality of locations to a motion planning system of the autonomous vehicle.