TESSE
Unity-based simulator to enable research in perception, mapping, learning, and robotics
Features
TESSE (Task Execution with Semantic Segmentation Environments) enables autonomy simulation across a diverse set of vehicles and use cases. Agents may be controlled via a Python API compatible with popular RL and robotics tools, such as ROS and OpenAI Gym. Sensor models expose data including semantic segmentation, depth, and inertial measurements.
Applications
TESSE has been used for applications including metric-semantic mapping, 3D dynamic scene graph generation, autonomous racing, and reinforcement learning for the GOSEEK Challenge.
Getting Started
You can find the TESSE source code on Github. See the README for instructions on using this for custom scenes.
Alternatively, you can use builds released for the GOSEEK Challenge.