KEYNOTE: SUMO and I – 25 years later...
summary
25 years of SUMO starts as a small DLR project that goes open source, gains TraCI (2007) as the control breakthrough and a user conference in 2013; current work adds a driver state device that injects perception noise and thresholds (Todosiev 1962) so SUMO can better match the empirical TTC distribution with its ±12 s peaks, whereas the default model overproduces large TTCs. Built features include a simple blackout crash mechanism and a 101‑car platoon test; the perception‑threshold model improves fit but remains only weakly empirically validated.
AI-written summary · about
chapters (29)
- 0:0025 years of SUMO intro
- 1:25SUMO’s early development
- 3:09Founding SUMO at DLR
- 4:55SUMO’s predecessors
- 6:19SUMO open source growth
- 9:53Driver modeling transition
- 10:13Microscopic driver challenges
- 12:01Physics-based driver modeling
- 13:02Driver behavior complexity
- 14:04Traffic safety research intro
- 15:07Modeling crashes in SUMO
- 16:30Safety surrogate measures
- 18:18Time to collision limits
- 20:02TTC distributions challenges
- 22:31Driver state device errors
- 24:14Perception thresholds control
- 25:16Testing driver state device
- 26:18Discontinuous traffic systems
- 27:21Platoon behavior analysis
- 28:271001st driver model
- 29:28Proposed model advantages
- 29:48Model testing real-world
- 30:51Action points analysis
- 32:58Empirical model support
- 34:03Action point distributions
- 35:52Conclusion on driver model
- 36:55Need for classical modeling?
- 37:57Machine learning future
- 38:40End of theory in traffic