R-00
Open questions we are working on.
JEISYN's product direction is shaped by a set of technical research themes. Each is framed as a question, with the methods of interest and the problems still to solve.
- Status vocabulary
- Implemented
Built and working in at least one internal configuration. - In development
Under active engineering; not yet feature-complete. - Experimental
Research prototypes and investigations; outcomes are uncertain. - Planned
On the roadmap; engineering has not started or is at the design stage. - Statuses appear only once confirmed by JEISYN.
Investigating visual-inertial odometry, LiDAR-based localisation and map-relative positioning for operation under bridges, in canyons, indoors and near interference.
Open questions
- Drift characterisation of visual-inertial odometry over long trajectories.
- Map-relative localisation against previously reconstructed terrain.
- Detection of GNSS degradation and safe fallback behaviours.
Methods of interest
- Visual-inertial odometry
- LiDAR odometry
- Factor-graph state estimation
Calibration and registration methods that keep heterogeneous sensor data spatially consistent across vehicles and missions.
Open questions
- Cross-modal calibration between thermal and visible cameras.
- Inter-vehicle map alignment without shared ground control.
- Uncertainty propagation into registered products.
Methods of interest
- Extrinsic calibration
- Point-cloud registration
- Bundle adjustment
Planning methods that treat coverage quality, not just distance, as the objective — and replan when a vehicle leaves or a priority changes.
Open questions
- Sensor-footprint-aware coverage metrics.
- Energy-constrained task allocation with return-to-charge cycles.
- Replanning triggered by coverage gaps in the world model.
Methods of interest
- Coverage path planning
- Auction-based allocation
- Receding-horizon replanning
Data structures and update rules for a shared, persistent model that records provenance, freshness and confidence for every layer.
Open questions
- Versioned spatial layers with provenance.
- Change detection between survey epochs.
- Efficient streaming of model updates to operators.
Methods of interest
- Multi-resolution spatial indexing
- Probabilistic occupancy
- Change detection
Georeferencing and interpolation of environmental readings, with explicit uncertainty, as first-class layers in the world model.
Open questions
- Spatial interpolation with uncertainty estimates.
- Calibration against ground reference stations.
- Combining thermal imagery with point measurements.
Methods of interest
- Gaussian-process interpolation
- Sensor calibration
- Data fusion
Bounded onboard decision-making — such as local replanning and obstacle response — within limits set and supervised by operators.
Open questions
- Onboard inference within power and thermal budgets.
- Explicit autonomy envelopes defined by operators.
- Safe behaviour on loss of link.
Methods of interest
- Embedded inference
- Behaviour trees
- Runtime assurance
Local-first networking and message prioritisation that keep the most important information flowing under constraint.
Open questions
- Priority-aware message scheduling.
- Store-and-forward for non-critical data products.
- Link-quality-aware planning.
Methods of interest
- Mesh networking
- DDS QoS policies
- Delay-tolerant networking
Interface and system design that keeps people in control: transparent intent, calibrated trust and clear authority.
Open questions
- Presenting plan intent and uncertainty to operators.
- Approval workflows for task assignment.
- Workload-aware alerting.
Methods of interest
- Human-factors evaluation
- Interface prototyping
- Scenario-based testing