Autonomous Vehicle Motion Forecasting with Diverse Joint Scenarios
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous vehicle systems struggle to accurately predict diverse future traffic scenarios, leading to inefficient motion planning and increased computational resource usage.
Innovation Solution
A motion forecasting and planning model that determines a compact set of diverse future scenarios by modeling the joint distribution of actor trajectories, using sensor data from LIDAR and HD maps, and employing a graph neural network to evaluate a diversity objective and generate a contingency plan.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Monte-Carlo sampling of latent variables is used to accurately predict future trajectories, then prediction accuracy is improved, but computational resources and time required increase prohibitively
Solution Approach 1:
The patent segments the prediction task by separating the sampling of shared latent variables from the generation of individual trajectory samples. Instead of performing full Monte-Carlo sampling for each trajectory, the system samples shared latent variables once and reuses them across multiple trajectory predictions, dividing the computational workload into manageable segments that can be processed efficiently
Solution Approach 2:
The system performs preliminary sampling of shared latent variables that encode unobserved scene dynamics before generating specific trajectory predictions. This preliminary action captures the essential scene context in advance, allowing subsequent trajectory generations to reuse these samples and avoid redundant computational effort
2Adaptability or versatility
If Monte-Carlo sampling is used to cover diverse future scenarios, then scenario diversity is improved, but the number of samples required becomes prohibitively large
Solution Approach 1:
The patent merges the sampling of shared latent variables with the trajectory generation process. By combining these previously separate operations into a unified approach where shared latent samples are generated once and reused across multiple trajectories, the system achieves diverse scenario coverage with fewer total samples
Solution Approach 2:
The shared latent variables serve multiple functions simultaneously: they encode unobserved scene dynamics, condition multiple trajectory predictions, and enable diverse scenario generation. This multi-functionality allows a single set of latent samples to support diverse trajectory predictions across different actors and time steps, reducing the total number of samples needed
3Device complexity
If marginal distribution prediction is used for each actor individually, then computational complexity is reduced, but scene consistency and interaction accuracy deteriorate
Solution Approach 1:
The patent introduces shared latent variables as an intermediary that connects individual actor predictions to the overall scene context. These latent variables act as a mediator that ensures scene consistency by conditioning all trajectory predictions on the same underlying scene dynamics, while still allowing efficient individual actor processing
Data Source
AI summary
Systems and methods are disclosed for motion forecasting and planning for autonomous vehicles. For example, a plurality of future traffic scenarios are determined by modeling a joint distribution of actor trajectories for a plurality of actors, as opposed to an approach that models actors individually. As another example, a diversity objective is evaluated that rewards sampling of the future traffic scenarios that require distinct reactions from the autonomous vehicle. An estimated probability for the plurality of future traffic scenarios can be determined and used to generate a contingency plan for motion of the autonomous vehicle. The contingency plan can include at least one initial short-term trajectory intended for immediate action of the AV and a plurality of subsequent long-term trajectories associated with the plurality of future traffic scenarios.