Actor Intention Forecasting for Autonomous Vehicle Trajectory Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face computational intensity in forecasting the intentions of actors in their environment, leading to resource inefficiencies in processing time and energy usage.
Innovation Solution
A method involving a perception system to detect actors, a forecasting system to generate and evaluate candidate intentions using data sets and machine learning models, and a motion planning system to select trajectories based on overall probabilities, refining predictions over cycles and considering environmental and kinematic states.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the AV performs comprehensive forecasting of actor intentions using multiple candidate intentions and probabilistic models, then the accuracy of intention prediction is improved, but the computational intensity and processing time increase
Solution Approach 1:
The forecasting system segments the intention prediction process into multiple independent candidate intentions, each evaluated separately with its own probabilistic model. This allows the system to maintain comprehensive analysis while organizing computations in manageable, parallelizable units that reduce overall processing time.
Solution Approach 2:
The system generates multiple candidate intentions beyond what is strictly necessary, then evaluates them with varying levels of detail. High-probability candidates receive more thorough analysis while low-probability ones are processed more quickly, allowing the system to achieve sufficient accuracy without exhaustive computation on all possibilities.
2Reliability
If the AV evaluates multiple candidate intentions with probabilistic models over multiple cycles, then the reliability of forecasting is improved, but the energy consumption increases
Solution Approach 1:
The system evaluates candidate intentions periodically over multiple cycles rather than continuously. Each cycle reassesses the probabilistic models and updates forecasts, maintaining reliability through repeated validation while reducing energy consumption by allowing periodic intervals between intensive computation cycles.
Solution Approach 2:
The system uses feedback from previous cycle evaluations to adjust the intensity of current cycle processing. High-probability candidate intentions that show consistent patterns across cycles receive continued attention, while those that diverge or resolve are processed more efficiently, optimizing the balance between reliability and energy usage.
3Adaptability or versatility
If the system generates and evaluates multiple candidate intentions for each cycle, then the completeness of forecasting is improved, but the device complexity increases
Solution Approach 1:
The forecasting system is segmented into distinct functional components: candidate intention generation, probabilistic model evaluation, and result aggregation. Each component handles a specific aspect of the comprehensive forecasting process, making the overall complex system manageable through modular design while maintaining complete coverage of possible intentions.
Solution Approach 2:
The probabilistic model evaluation framework serves multiple functions simultaneously: it generates candidate intentions, evaluates their likelihood, tracks persistence across cycles, and determines overall probabilities. This multi-functionality reduces the need for separate specialized systems for each forecasting task, managing complexity while maintaining completeness.
Data Source
AI summary
Systems and methods for controlling an autonomous vehicle (AV). The methods comprise: generating candidate intentions of an actor based on a detected action of the actor and a classification associated with the actor; determining an overall probability for each candidate intention based on at least a persistence of the candidate intention over a non-interrupted sequence of cycles (where each cycle represents a time period over which the actor was sensed by a sensor); selecting candidate intention(s) based on the overall probabilities; forecasting a subsequent future intention that the actor may have after reaching a goal defined by the candidate intention(s) which was(were) selected; obtaining an actor trajectory that is consistent with the candidate intention(s) which was(were) selected and the subsequent future intention; and using the actor trajectory to influence a selected trajectory for AV.