Jupedsim

Release Notes

This release introduces a new model, deprecations, and bug fixes.

Anticipation Velocity Model (AVM)

The AVM models pedestrian anticipation in three phases:
1. **Perception** of the current situation.
2. **Prediction** of future movement.
3. **Strategy selection** leading to action.

By incorporating these steps, the AVM quantitatively reproduces bidirectional pedestrian flow. It accounts for:
- Anticipation of changes in neighboring pedestrians' positions.
- The strategy of following others' movement.

More details on the AVM: [Anticipation Velocity Model](https://pedestriandynamics.org/models/anticipation_velocity_model/)

Deprecations

CamelCase Naming Deprecated

In our last releases several properties slipped into the release that did not follow PEP8 naming conventions, this has been corrected and the camel case style names have been deprecated.

`v0` renamed to `desired_speed`

This update also deprecates `v0` in all models and replaces it with the much clearer `desired_speed`.

`e0` renamed to `desired_direction`

This update also deprecates `e0` in all models and replaces it with the much clearer `desired_direction`.

Bug fixes

- Fixed [1324](https://github.com/PedestrianDynamics/jupedsim/issues/1324)
- Fixed [1413](https://github.com/PedestrianDynamics/jupedsim/issues/1413)
- Fixed [1449](https://github.com/PedestrianDynamics/jupedsim/issues/1449)

What's Changed

* Fixed WaitingSet: In specific cases the wrong waiting position was returned, this should work as expected now.
* Fixed serialisation issue: The serialisation is writing out a bounding box over the union of all geometries, min/max values had been swapped.

This release contains two new features

New Model - Social Force Model

JuPedSim now implements another microscopic model, the `Social Force Model` as described by Helbing, D., Farkas, I., Vicsek, T. (2000). Simulating dynamical features of escape panic.

New route planning method - Direct Steering

Direct Steering allows the user to set each agents target individually and at any time. Agents will then use normal way findig to navigate to the set position. This should allow for rapid prototyping of high level agent behaviour such as waiting or queueing. The intention here is that user will be able to use their own logic written in python to do route planning. While this needs extra coding and might be runtime intensive it allows full flexibility.

What's Changed
* The SQL output does now properly include all geometry variants when geometries are switched at runtime.

What's Changed
* Add possibility to switch geometry at runtime.
* Add new iteration of the Collision Free Speed Model (CollisionFreeSpeedModelV2), in this update formerly global repulsion parameters have become agent specific parameters. Thus allowing to model individual agents to react with different strength to geometry / neighbouring agents.

This is a bugfix release:

* Fix typo in documentation
* Address address sanitiser finding when freeing memory for journeys