This directory represents the codebase for ongoing research on temporal convolutions for multi-step quadrotor motion prediction. This includes all code necessary to develop fully-convolutional Temporal Convolutional Networks (TCNs), hybrid models, and physics-based numerical simulations for robotic system modeling. It also includes all code used to train and test predictive models and generate any published results.
This includes the following files:
- data_loader.py: Generate custom PyTorch datasets for quadrotor multistep motion prediction
- End2EndNet.py: Build and train End2EndNet for robotic system modeling
- HybridTCN.py: Build and train TCN hybrid models for quadrotor modeling
- PhysicsModel.py: Build and simulate physics-based quadrotor models
- SystemID.py: Perform system identification for physics-based quadrotor models
- neuralnet_eval.py: Evaluate neural network robotic system predictive models over multiple steps
- physicsmodel_eval.py: Evaluate physics-based robotic system predictive models over multiple steps
- prediction_sim.py: Simulate robotic system motion and predictive models over trajectory samples
- dataset_stats.py: Calculate dataset statistics