GOSEEK Challenge @ ICRA 2020
An autonomous treasure hunt challenge.
Overview
Hosted in conjunction with the Perception, Action, Learning Workshop, the GOSEEK challenge consists in creating an agent that combines advanced perception and high-level decision-making to search for objects placed within complex indoor environments from a Unity-based simulator. Simply put: like PACMAN, but in a realistic scene and with realistic perception capabilities. The competition page provides further detail on the problem definition, logistics, and participation.
Getting Started
We provide a baseline policy consisting on a CNN-LSTM trained with Proximal Policy Optimization (PPO). You can check out the training notebook to get started.
Submission & Results
The contest was hosted on the EvalAI platform, where participants submit solutions, via docker images, for scoring. Checkout out our page on EvalAI for results!