Crowd Navigation Using Interacting Gaussian Processes for Intent Handling

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Classical congestion navigation algorithms fail to optimize intent or flexibility, particularly in multi-faceted environments, leading to host-agent agreement failures and requiring large data sets or handcrafted objective functions.

Innovation Solution

A system utilizing interacting Gaussian processes, specifically zero free-parameter Gaussian processes (zpIGP), to model host-agent interactions, optimizing joint predictive distributions and obstacle avoidance functions, which recalculate means, covariances, and mixture weights dynamically to achieve global optimality with convex optimizations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If classical congestion navigation algorithms use deep learning approaches with handcrafted objective functions, then navigation optimization may be achieved, but the system requires human practitioner intervention and fails to properly handle multi-faceted intent

Engineering Contradiction:

Improvehost-agent agreementVSAvoidmulti-faceted intent handling

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The system employs self-service through automatic objective function generation using Gaussian process regression. Instead of requiring human practitioners to handcraft objective functions, the system automatically learns and generates appropriate objective functions from data, enabling it to handle multi-faceted intent without human intervention while maintaining reliable host-agent agreement

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system applies parameter changes by transforming the objective function from a fixed handcrafted form to a dynamic form generated by Gaussian process regression. This allows the objective function parameters to adapt based on learned patterns from data, enabling proper handling of multi-faceted intent while maintaining optimization reliability

Inventive Principle:

Principle #35Parameter changes

2Adaptability or versatility

If classical congestion navigation algorithms use data driven representations, then representations may be learned, but large amounts of data are required for acquisition

Engineering Contradiction:

Improvedata driven representationVSAvoiddata quantity

Core Design Contradiction:

Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The system applies preliminary action by pre-defining the Gaussian process framework and its probabilistic structure before data acquisition. This preliminary setup allows the system to learn effective representations with smaller data quantities, as the Gaussian process prior provides structured assumptions that reduce the data needed compared to purely data-driven approaches

Inventive Principle:

Principle #10Preliminary action

3Productivity

If classical congestion navigation algorithms minimize joint cost with sampled models, then navigation paths may be optimized, but intent optimization and flexibility are failed

Engineering Contradiction:

Improvenavigation optimizationVSAvoidintent and flexibility

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The system substitutes the mechanical sampling and joint cost minimization approach with a probabilistic Gaussian process framework. This replacement enables the system to simultaneously optimize navigation paths and intent/flexibility by modeling uncertainties and preferences probabilistically, rather than through deterministic sampled models

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

4Device complexity

If interacting Gaussian processes are used to model host-agent interactions, then computational complexity is reduced and global optimality is achieved, but the system requires dynamic recalculation of means, covariances, and mixture weights

Engineering Contradiction:

Improvecomputational complexityVSAvoiddynamic recalculation

Core Design Contradiction:

Device complexityVSExtent of automation

Solution Approach 1:

The system applies dynamics by implementing dynamic recalculation of Gaussian process parameters (means, covariances, and mixture weights) as agents interact and new observations are made. This dynamic adaptation enables the system to maintain accurate probabilistic models of host-agent interactions while reducing overall computational complexity through the structured Gaussian process framework

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS11787053B2Systems and methods for utilizing interacting Gaussian mixture models for crowd navigation

Publication Date: 2023.10.17 HONDA MOTOR CO LTD

AI summary

Systems and methods for utilizing interactive Gaussian processes for crowd navigation are provided. In one embodiment, a system for a crowd navigation of a host is provided. The system includes a processor, a statistical module, and a model module. The processor receives sensor data. The statistical module identifies a number of agents in a physical environment based on the sensor data. The statistical module further calculates a set of Gaussian processes. The set of Gaussian processes includes a Gaussian Process for each agent of the number of agents. The statistical module further determines an objective function based on an intent and a flexibility. The model module generates a model of the number of agents by applying the objective function to the set of Gaussian processes. The model includes a convex configuration of the number of agents in the physical environment.