This project is for the evaluation of the Event-Triggered Generalized Outcome Assessment (ET-GOA) algorithm using both a simulated and live robot.
ET-GOA leverages the Generalized Outcome Assessment (GOA) and Model Quality Assessment (MQA) metrics from Factorized Machine Self-Confidence (FaMSeC) to enable an autonomous robot to understand when and how its competency changes during task execution. This project extends our previous work:
Generalizing Competency Self-Assessment for Autonomous Vehicles Using Deep Reinforcement Learning
Dynamic Competency Self-Assessment for Autonomous Agents
We published this article:
Event-triggered robot self-assessment to aid in autonomy adjustment
Ubuntu 20.04
Python: 3.8.10
ROS: noetic
Webots: R2023a
Start some ROS environment if this is for simulation (I use the base Jackal):
$ roslaunch jackal_gazebo empty_world.launch
Start the user interface:
$ python3 ./interface/ros_ui_main.py
The interface should be started. Telemetry should be flowing. Entering a goal in the (x,y) boxes will update the goal. Plans are generated using RRT. Auto starts autonomous navigation, stop stops navigation, and telop waits for telop commands. Robot competency can be seen on the right panel. Below are several buttons to turn on/off different functions and end the simulation.