Autonomous Vehicle Motion Forecasting With Joint Traffic Scenarios
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous vehicle forecasting systems struggle to accurately predict diverse future traffic scenarios due to the limitations of modeling individual agent trajectories, leading to inefficient resource usage and high latency, as they fail to account for interactions among multiple agents in a scene.
Innovation Solution
A system that models the joint distribution of actor trajectories for multiple agents using a graph neural network (GNN) to determine a compact set of diverse future scenarios, optimizing a planner cost function that includes comfort, motion rules, and route considerations, while reducing the number of required samples.
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 the number of samples required becomes prohibitively large, increasing computational resources and latency
Solution Approach 1:
The patent merges the prediction of multiple agents' trajectories into a single joint distribution model. Instead of independently sampling trajectories for each agent (which requires many samples to capture interactions), the system samples from a unified joint distribution that inherently encodes agent-agent interactions, reducing the number of samples needed while maintaining accuracy.
Solution Approach 2:
The patent transitions from marginal distribution modeling (individual agent trajectories) to joint distribution modeling (collective agent behavior). This dimensional shift from individual to collective space allows the system to capture interaction effects without requiring proportional increases in sampling effort.
2Productivity
If marginal distributions of individual agent trajectories are modeled, then computational resources are reduced, but scene consistency and interaction accuracy deteriorate
Solution Approach 1:
The system combines individual agent trajectory predictions into a joint distribution that preserves scene consistency. By modeling agents collectively rather than independently, the system maintains realistic interaction patterns while avoiding the computational burden of extensive sampling.
3Adaptability or versatility
If a large number of future traffic scenarios are sampled to cover diverse possibilities, then scenario coverage is improved, but computational latency increases
Solution Approach 1:
The patent combines diverse scenario generation with joint distribution sampling. A single sample from the joint distribution inherently produces a consistent set of trajectories for multiple agents, capturing their interactions and producing diverse yet coherent future scenarios without requiring multiple independent sampling passes.
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.