------------------- Add: Login feature and UI required for authentication in multi-agent missions Fix: Render size set correctly, avoiding aspect ratio distortion. Add: Support for multi-agent missions. New sample: MultiMaze.py
0.11.0
------------------- Change: All project code renamed to 'Malmo'. Add: New ALEAgentHost, to use the Atari Learning Environment as a back-end instead of Minecraft. Add: Chat commands and observations now supported. Fix: MissionSpec::observeGrid and ::observeDistance were broken.
0.10.4
------------------- Fix: discrete movement had interpolation when rendering and offsetting in the Mod when bumping against walls. Fix: tabular_q_learning.py was using deprecated API.
0.10.3
------------------- Change: replaced MissionSpec::useDiscreteActions() with allowContinuousMovementCommand() etc. to control allow- and deny-lists through the API. Fix: 64-bit build now supported on Windows. Fix: Human Action Component now supports discrete actions again. Add: New visualization tool for heatmaps of agent movement.
0.10.2
------------------- Fix: ffmpeg now a runtime dependency instead of an installation requirement. Fix: pitch and yaw speed now render-speed independent.
0.10.1
------------------- Add: Tutorial pdf and scripts Fix: DefaultWorldGenerator wasn't generating if a flatworld was already running. Fix: ObservationFromGrid was never returning observations. Fix: Hotbar keys were broken for both agent and human. Fix: Initial yaw and pitch were being stomped on by ContinuousCommandHandler. Fix: The order of draw objects in the DrawingDecorator is now preserved. Fix: Command handler modifiers now called "allow-list" and "deny-list". Add: AI/Human toggle now bound to enter key and state is displayed in info text. Fix: Mouse can no longer move the AI in the gap between missions. Fix: AI/Human toggling of control fixed (no longer need to press toggle key twice). Fix: Empty JSON observations filtered out. Add: Tutorial in Python_Examples.