Duo

Duo is the companion code for the RSS workshop paper "Exploring Active Human Goal Inference in Shared Autonomy and Autonomous Driving" from the InterACT lab.

I called it duo because all of our examples explore inference between two goals.

Running an Experiment

Write an <experiement-name>.json file in experiments/ (see experiments/ for examples). Optionall write an <experiment-name.py` companion python file, with functions for plotting the results (c.f. any of experiements/*.py).

The experiment can be run:

pyhonn exp.py experiments/<experiment-name>.json

Organization of Code

• human.py contains our human models
• robot.py contains our robot models (algorithms)
• beliefs.py contain the observation model, and bayesian inference utilities
• sim.py simulates the interaction between a robot and human given initial config
• exp.py takes an initial configuration (including algorithm for robot, model of robot, start, goals) and runs it, optinally creating plots of the results, whihch are defined in a companion file.