Multi-Actor Trajectory Prediction Using Confidence Maps and Vector Fields
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems for autonomous vehicles are limited in predicting future trajectories of multiple actors in a multi-actor environment, as they typically compute individual trajectories sequentially, leading to inefficiencies and inaccurate predictions, especially in real-time applications.
Innovation Solution
A deep neural network (DNN), such as a recurrent neural network (RNN), is used to predict future locations of actors by leveraging past locations, map information, wait conditions, and environmental data, allowing for simultaneous computation of trajectories for multiple actors, and generating confidence maps and vector fields to track actors over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional systems compute individual trajectories sequentially for each actor, then the processing burden is manageable, but the runtime becomes ineffective for real-time deployment
Solution Approach 1:
The patent merges the trajectory prediction process for multiple actors into a single unified neural network computation. Instead of sequentially processing each actor individually, the system combines all actor trajectory predictions into one parallel computation pass through the neural network, achieving real-time performance while maintaining accuracy.
2Measurement precision
If conventional systems use individual computations for each actor based on high-level features, then the approach is computationally feasible, but the accuracy of multi-actor trajectory predictions deteriorates
Solution Approach 1:
The system combines multiple actor trajectory predictions into a single neural network forward pass, where the network simultaneously processes all actors and outputs their trajectories in one computation. This unified approach captures inter-actor relationships and environmental context that individual computations miss, improving accuracy while achieving real-time speeds.
Solution Approach 2:
The neural network is designed as a universal predictor that handles multiple actors of different types (pedestrians, vehicles, cyclists) simultaneously using the same computational framework. The network accepts multi-actor input data and produces coordinated trajectory predictions for all actors in the scene, rather than requiring separate specialized computations for each actor type.
3Adaptability or versatility
If conventional systems predict many possible trajectories using variational autoencoders, then potential future paths are explored, but the predictions do not directly correspond to actual predicted future trajectories
Solution Approach 1:
The system extracts the most relevant trajectory information directly from neural network output layers that are specifically designed to predict actual future positions. Rather than generating multiple possible trajectories and then selecting among them, the network directly outputs the predicted future trajectory coordinates through dedicated output neurons, providing accurate and actionable predictions.
Data Source
AI summary
In various examples, past location information corresponding to actors in an environment and map information may be applied to a deep neural network (DNN)—such as a recurrent neural network (RNN)—trained to compute information corresponding to future trajectories of the actors. The output of the DNN may include, for each future time slice the DNN is trained to predict, a confidence map representing a confidence for each pixel that an actor is present and a vector field representing locations of actors in confidence maps for prior time slices. The vector fields may thus be used to track an object through confidence maps for each future time slice to generate a predicted future trajectory for each actor. The predicted future trajectories, in addition to tracked past trajectories, may be used to generate full trajectories for the actors that may aid an ego-vehicle in navigating the environment.