This repository hosts InfoFuseSched, a value-aware world-state data-flow
scheduler for human-machine-object fusion scenarios.
Instead of scheduling isolated tasks, InfoFuseSched models sensor frames, robot
poses, human requests and safety events as a continuously-updated world-state
graph, and schedules the refresh, reuse and distribution of shared state views:
multiple tasks that depend on the same view are served by one refresh, similar
inputs reuse a semantic cache, local changes trigger incremental updates, and
data is shipped as raw / feature / edge-local / cached results according to a
Value-of-Information vs cost trade-off.
The project runtime, benchmark, scenarios, tests, deployment and documentation lives under infofuse-sched/.
Quickstart
cd infofuse-sched
bash scripts/setup.sh # machine-safe local virtualenv
make check # formatter, linter and tests
make build # Python package build
make test # unit + smoke tests
make experiments # run benchmarks and refresh dashboards
make serve # preview dashboards on http://127.0.0.1:18084
See infofuse-sched/README.md for the architecture, algorithms, experiments and machine-safety notes.
openharmony
This repository hosts InfoFuseSched, a value-aware world-state data-flow scheduler for human-machine-object fusion scenarios.
Instead of scheduling isolated tasks, InfoFuseSched models sensor frames, robot poses, human requests and safety events as a continuously-updated world-state graph, and schedules the refresh, reuse and distribution of shared state views: multiple tasks that depend on the same view are served by one refresh, similar inputs reuse a semantic cache, local changes trigger incremental updates, and data is shipped as raw / feature / edge-local / cached results according to a Value-of-Information vs cost trade-off.
The project runtime, benchmark, scenarios, tests, deployment and documentation lives under
infofuse-sched/.Quickstart
See
infofuse-sched/README.mdfor the architecture, algorithms, experiments and machine-safety notes.