Quantum Tracking Method for Target Data Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional target tracking methods, such as Multiple Hypothesis Tracking (MHT), face challenges in accurately associating observations with tracks, especially when closely spaced targets produce conflicts and multiple observations fall within a track's gate, leading to inefficiencies in data association and track initiation.
Innovation Solution
A hybrid quantum/classical computer method processes sensor data by constructing a graph of sensed plots as vertices and edges, associating binary variables to edges to favor geometrically relevant patterns, and optimizing a cost function using a quantum processing system to determine the ground state of a Hamiltonian encoding the solutions, thereby improving data association and track management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional Multiple Hypothesis Tracking (MHT) methods are used to handle data association, then the system can manage multiple targets, but the accuracy deteriorates when closely spaced targets produce conflicts and multiple observations fall within a track's gate
Solution Approach 1:
The patent transforms the data association problem from traditional probabilistic methods to a quantum optimization framework by changing the mathematical representation parameters. The cost function is encoded into a Hamiltonian operator where binary variables represent association decisions, and quantum mechanics principles (superposition, entanglement) enable simultaneous evaluation of multiple hypotheses, resolving the accuracy issue with closely spaced targets
Solution Approach 2:
The patent replaces the classical computational mechanics of MHT with quantum mechanical processes. Instead of sequentially evaluating hypotheses through classical algorithms, the system uses quantum state evolution, Hamiltonian dynamics, and quantum measurement to determine optimal data associations, achieving superior precision in distinguishing closely spaced targets
2Adaptability or versatility
If traditional MHT methods process observations through multiple hypotheses, then the system can handle complex tracking scenarios, but the computational complexity and processing time increase significantly
Solution Approach 1:
The patent merges the hypothesis generation and evaluation steps into a unified quantum optimization process. By encoding multiple hypotheses into a single quantum state superposition and using Hamiltonian evolution to simultaneously process all associations, the system achieves versatility in handling complex scenarios while reducing computational complexity through quantum parallelism
Solution Approach 2:
The transformation from classical cost minimization to quantum energy minimization changes the computational parameters fundamentally. The cost function becomes a Hamiltonian operator, and the optimization process leverages quantum tunneling and interference effects to escape local minima, reducing the computational burden of evaluating multiple hypotheses
3Reliability
If traditional gating methods are used to filter observations, then the system can reduce false associations, but the ability to distinguish true targets from false alarms deteriorates when targets are closely spaced
Solution Approach 1:
The patent changes the discrimination parameter from classical gate-based spatial filtering to quantum cost function evaluation. The Hamiltonian encoding incorporates geometric relationships and motion dynamics, enabling the system to distinguish true targets from false alarms by evaluating the overall consistency of associations rather than relying solely on spatial proximity thresholds
Data Source
AI summary
The present disclosure provides a hybrid quantum/classical computer implemented method useful in tracking one or more targets in a field of view. The method comprises: receiving positioning data and or not doppler data from a sensing system scanning a field of view containing one or more targets over successive sensing times (t0, . . . tk) thereby collecting over time a batch (Bk) of scans (S1, . . . , Sk) of sensed plots. The method further comprises, for a current batch (Bk) of scans (S1, . . . , Sk) of sensed plots: processing a graph (G) containing said sensed plots as vertices (V) and possible connections between pairs of plots associated with consecutive scans as edges (E); associating a binary variable to each edge of the graph (G); processing a cost function of the binary variables so as to favor geometrically relevant edge patterns for the targets being tracked; optimizing said cost function using a quantum processing system by determining the ground state of a Hamiltonian encoding the solutions of said cost function in its eigenvalues; and processing one or more tracks between plots of successive scans when said plots belong to an edge associated with a binary variable solving said optimization.