Sup-HMM
Taming Volatile Trajectories: a Bayesian-optimized HMM for location-based service data.
Mobility representation engine
NovaMatch.
From GPS observations to transportation network representations.
A trajectory processing framework for recovering mobility paths from noisy observations and translating them across road and planning networks.
Richmond · sample trajectoryNetwork path recovered
INPUTObserve
Sparse, noisy GPS trajectories
→01Recover
Network-constrained mobility paths
→02Translate
Road and planning representations
Map matching
Recover paths on
any digital network.
NovaMatch recovers network-constrained mobility paths from sparse and noisy observations across OSM-derived or user-defined digital networks.
Custom networks
OSM or user-defined geometry.
Sparse observations
Recover continuity from incomplete trajectories.
Scalable processing
For large mobility datasets and experiments.
Cross-network translation
One trip.
Different network logic.
Preserve mobility semantics when road trajectories meet abstract planning networks.
Observed mobilityRoad trajectory
N
NovaMatchsemantic translation→Planning representationSame space · abstract logic
Scaling NovaMatch · 03 / 03
Process less.
Preserve more.
PM-Tree asks a deeper question: how much trajectory information is actually needed to preserve path recoverability?
Scientific lens: trajectory information sufficiency.
System role: structural representation and acceleration.
System role: structural representation and acceleration.
PM-Tree · representative result
100%original trajectory
~3%retained points
~0.96path F1
Interactive prototype
See NovaMatch work.
Submit observations and see the recovered network path.
Ready · 8 observations loaded
Research behind NovaMatch
A coherent system,
grounded in research.
PM-Tree
How Much of a Trajectory Is Needed? Priority-guided hierarchical trajectory representation.
Selective Translation
Mapping observed mobility into abstract links, connectors, and planning-network semantics.