Problem
Remote routing requires internet connectivity and adds latency. DeFlock already supports offline map areas — offline routing would complete the offline experience and remove server dependency.
Goal
Investigate routing engines that could run on-device in DeFlock with camera avoidance support and downloadable regional data.
Research So Far — CoMaps/Organic Maps Engine
One option explored is the CoMaps routing engine (Apache 2.0, C++, fork of Organic Maps/MAPS.ME):
- Custom bidirectional A* algorithm in C++
- On-device routing with pre-downloaded
.mwm regional data
- Existing penalty system (7 types × 4 vehicle classes) with extensible architecture
EdgeEstimator::CalcSegmentWeight() is the key injection point for camera avoidance penalties- Road avoidance already exists (tolls, unpaved, motorways, ferries) via bitmask system
- Geographic avoidance zones are architecturally feasible: extend
EdgeEstimator, inject camera database, query proximity during weight calculation
- No existing Flutter bindings — would need platform channels (C++ via JNI on Android, Swift bridge on iOS) or a Dart rewrite
Other Directions to Explore
- Valhalla (C++, MIT license) — supports custom costing profiles, more feature-rich avoidance
- OSRM (C++, BSD license) — proven offline routing, simpler model
- Pure Dart implementation — rewrite A* routing in Dart, no FFI needed, simpler integration but needs custom map data preprocessing
- GraphHopper Mobile — Java-based, Android-native, may have iOS limitations
- flutter_osm_plugin — existing Flutter package with offline routing support
Key Design Decisions Needed
- C++ via FFI vs. pure Dart rewrite
- Map data format (MWM? Custom binary? OSM PBF preprocessing?)
- Downloaded regions model (user picks state/metro area, ~100–500 MB per region)
- How to inject camera locations into the routing graph (preprocessing vs. runtime penalty)
- How to keep camera data fresh for offline use
Relationship to Remote Routing
This is exploratory — the short-term remote routing (#118) addresses the immediate need while this investigation proceeds.
Category
Routing · Research · Enhancement
Problem
Remote routing requires internet connectivity and adds latency. DeFlock already supports offline map areas — offline routing would complete the offline experience and remove server dependency.
Goal
Investigate routing engines that could run on-device in DeFlock with camera avoidance support and downloadable regional data.
Research So Far — CoMaps/Organic Maps Engine
One option explored is the CoMaps routing engine (Apache 2.0, C++, fork of Organic Maps/MAPS.ME):
.mwmregional dataEdgeEstimator::CalcSegmentWeight()is the key injection point for camera avoidance penaltiesEdgeEstimator, inject camera database, query proximity during weight calculationOther Directions to Explore
Key Design Decisions Needed
Relationship to Remote Routing
This is exploratory — the short-term remote routing (#118) addresses the immediate need while this investigation proceeds.
Category
Routing · Research · Enhancement