SubT-NCTU

Competitions

Urban Circuit

This work is the initiative of Team NCTU's participations to the DARPA Subterranean (SubT) Challenge, which is a US$2 million prize competition funded by the Defense Advanced Research Projects Agency (DARPA). DARPA is a prominent research organization of the United States Department of Defence, aiming to explore new approaches of mobility, perception, autonomy, and communication to navigate, map, and search underground environments with hazards, often too high-risk for personnel to enter. The participated teams include top universities and organization: NASA JPL, MIT, ETH Zurich, UC Berkeley, Cal Tech, KAIST etc. Team NCTU ranked the 8th among the 10 teams in the Urban Circuit in 2020.

Tunnel Circuit

Team NCTU ranked the 7th among the 11 teams in the Tunnel Circuit in 2019 (2019-08-22). We completed the competition in the challenging underground environments with limited communication and rough terrain. We managed to make our robot to autonomously navigate and score, resulting in the 7th place among the 11 remarkable teams. Team NCTU will participate the Urban Circuit in Feb., 2020.

DARPA News: Rolling, Walking, Flying, and Floating, SubT Challenge Teams Traverse the Tunnel Circuit>

Pittsburgh's NPR News: Wheels, Drones And Rescue Randy: DARPA Robotics Competition Puts Mine Rescue To The Test

Early Preparations

The Team NCTU for DARPA Subterranean Challenge is composed of graduate and undergraduate students from Institute of Electrical and Control Engineering, Department of Electrical and Computer Engineering, Mechanical Engineering, and the Robotics Institute, supervised by Dr. Hsueh-Cheng Nick Wang and 4 other faculty advisors. We hope to put together the efforts from talented and dedicated students here, and develop cutting-edge intelligent systems that solve the challenges encountering subterranean environments, where the robots must deal with rough terrain, degrading sensing, and severe communication. Our members are trained from the Duckietown platform, and some are the core member in the RobotX Competition 2018, who won the 5th place and the Best Single Day Award. There are in total 23 members in our team, including 2 engineers from our industrial partners, Acer Foundation and K-Best. The members in NCTU are developing the aspects of autonomy, perception, and mobility, whereas our industrial partners provide communication solutions. We consider the participations of the DARPA SubT Challenge both for research and education efforts. We establish internship programs to involve undergraduate students. We wish the team is student-centered, owned and run, with the help of the faculty advisors. We plan to share the efforts to the communities, publish results in top robotics conferences, and later adapt our works for real-world applications such as tunnel inspection in civil engineering and nuclear decommissioning.

Our approaches are deeply inspired by our recent work in the Duckietown & AI Driving Olympics 2018-2019, and the Maritime RobotX Competition 2018. Our long-term flight, collision-safe aerial blimp robot Duckiefloat is adapted from the Duckebot in the Duckietown platform, the miniaturized testbed to develop autonomy education and research of a fleet of robots. Duckiefloat mimics the lane following in Duckietown for tunnel following, and altitude controls from our underwater robot developed in the RobotX Competition. The Anchorball launcher on the UGV Husky is adapted from the one on our WAM-V (wave adaptive modular vehicle) in RobotX. We continue developing a revised Duckietown, including multiple robots with similar setups to Duckiefloat, artifacts, and tunnel-like environments, as a testbed for DARPA SubT Challenge.

Publications

J.-T. Huang, C.-L. Lu, P.-K. Chang, C.-I Huang, C.-C. Hsu, Z. L. Ewe, P.-J. Huang, and H.-C. Wang "Cross-Modal Contrastive Learning of Representations for Navigation using Lightweight, Low-Cost Millimeter Wave Radar for Adverse Environmental Conditions" in IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 3333-3340, April 2021, doi: 10.1109/LRA.2021.3062011. Link

Deep reinforcement learning (RL), where the agent learns from mistakes, has been successfully applied to a variety of tasks. With the aim of learning collision-free policies for unmanned vehicles, deep RL has been used for training with various types of data, such as colored images, depth images, and LiDAR point clouds, without the use of classic map--localize--plan approaches. However, existing methods are limited by their reliance on cameras and LiDAR devices, which have degraded sensing under adverse environmental conditions (e.g., smoky environments). In response, we propose the use of single-chip millimeter-wave (mmWave) radar, which is lightweight and inexpensive, for learning-based autonomous navigation. However, because mmWave radar signals are often noisy and sparse, we propose a cross-modal contrastive learning for representation (CM-CLR) method that maximizes the agreement between mmWave radar data and LiDAR data in the training stage. We evaluated our method in real-world robot compared with 1) a method with two separate networks using cross-modal generative reconstruction and an RL policy and 2) a baseline RL policy without cross-modal representation. Our proposed end-to-end deep RL policy with contrastive learning successfully navigated the robot through smoke-filled maze environments and achieved better performance compared with generative reconstruction methods, in which noisy artifact walls or obstacles were produced. All pretrained models and hardware settings are open access for reproducing this study and can be obtained at https://arg-nctu.github.io/projects/deeprl-mmWave.html.

Lu, C.-L., Huang, C.-I, Huang, J.-T., Hsu, C.-C., Chang, P.-K., Ewe, Z., Huang, P.-J., Huang, Y.-Y., Huang, S.-C., Li, P.-L., Wang, B.-H., Yim, L.-S., Huang, S.-W., Liu, Z.-Y., Bai, M.-H., Wang, H.-C. (2021). A Heterogeneous Unmanned Ground Vehicle and Blimp Robot Team for Search and Rescue using Data-driven Autonomy and Communication-aware Navigation, Field Robotics. (In revision) Link

This paper describes the architecture and implementation of a heterogeneous team comprising unmanned ground vehicles and blimp robots capable of navigating unknown subterranean environments for search and rescue missions. The ground vehicles are equipped with a range of sensors for accurate perception, localization, and mapping. The blimps feature a long flight duration and collision tolerance when traversing uneven terrain. The design of the system was meant to satisfy the requirements of the DARPA Subterranean Challenge in terms of perception capability and autonomy. To facilitate navigation through smoke-filled spaces, we employed novel millimeter wave radar to enable cross-modal representations for integration via deep reinforcement learning. The autonomy of the proposed scheme was augmented using simulations to train deep neural networks, thereby allowing the system to perform sequential decision-making for collision avoidance and navigation toward a specific goal. The navigation system was evaluated in the DARPA SubT Urban Circuit, and quantitative localization results and recovery strategy from failures was discussed. The proposed communication system comprises mesh WiFi with XBee and UWB communication modules as well as spherical nodes that can be shot out like a cannonball and miniature cars deployed as mobile nodes. The propagation and radio signal strength index of various modules were evaluated using data collected during field tests in order to overcome the uncertainties of subterranean environments, including non-line-of-sight propagation, multipath propagation, and fading reception. We also discuss the lessons learned during this project and reflect on future plans.