The summary video shows the simulation and real-world experimental demonstrations of our SOGMP/SOGMP++ predictors in crowded dynamic scenes:
1. Three GIFs show the occupancy grid map prediction comparison results (0.5s, or 5 time steps into the future) of our proposed SOGMP++ and SOGMP algorithms along with the ConvLSTM, PhyDNet, DeepTracking, SOGMP_NEMC baselines. Each GIF shows the results from a random sequence from a dataset, each of which has data from a different robot model.
2. Simulation demonstrations of the DWA and DWA-PU control policies in the indoor lobby.
3. Hardware demonstration of the DWA-PU control policy in the indoor hallway.
