Advanced Robotics Centre, National University of Singapore
For more info, contact Marcelo H Ang Jr (mpeangh@nus.edu.sg)

Robotics Capabilities

Compliant Motion and Force Control

Video: A ship welding robotic system. A powered arm is walk-through programmed the required motions and welding actions. We custom designed our own user interface for easy to use programming. The welding tool is mounted on the end-effector equipped with a six axis force torque sensor. The controller is a SUN HyperSpark Station on the VME bus of the Reis Robotic Manipulator.

Video: Our 3-DOF mini-manipulator (parallel manipulator configuration) in force control mode. Motion stops as soon as manipulator touches edge of paper. Note that the paper does not bend.
Video: Our 3-DOF Mini-manipulator in impedance control mode  

Video: Zero Moment Control - following an unknown surface. Moment around the x and y axes of the contact task frame is controlled to be zero allows the robot to automatically turn it's end-effector to follow the unknown surface.

Mpg video file: Canopy Polishing using Compliant Motion with Base Motion as Disturbance

Video: a Kuka IIWA following a unknown surface with a sharp discountinuity (corner) using Impedance Control. A force normal to the surface is commnanded togethet with a motion trajectory parallel to the surface.

Mpg video file: Canopy polishing using compliant motion across wrist singularity

Mpg video file: Singularity handling on the PUMA 560 robot

Video: Impact is handled by damping control

Human-Robot Systems

Video: A brain-controlled wheelchair - controlled by EMG signals. The "oddball" paradigm is deployed -> to allow the user to select the destination desired from a 3 x 3 table of possible destinations that are flashing. Since we control the flashing, the destination desired is matched with a synchronized EMG spike.

Video: Dorothy - A social robot - video shows the different possible movements allowing dorothy to express different emotions

Video: A head mounted display with head tracking - allowing a user to see what a remote robot is seeing.

Video: Leaning human motion skill of object handover. Robot learns when to natually let go.

Soft Robotics

Video: Soft Robotic Gripper made of fabric rotary actuators that are air activated.
Video: Soft Robotic Finger made of silicone with embedded cotton threads (orange) at varying densities (threads/mm). this allowing designed curve response when air-activated.
Video: Worm-like robot using Dielectric Elastomer Acrtuator (3m VHB Tape)Video: Air activated silicone fingers - airchannels insde the silicon material allows the finger to bend naturally.
Video: Robotic gripper with 3 soft fingers (air activated silicone) on a Kuka IIWA.
Video: Gripper releasing the object - see how soft the grippers are

Mobility

An autonomous wheelchair (based on the Whill platform) in Singapore National Gallery

whill-national-gallery-2019-07-01-xx.mp4

 

Video: End-to-End Deep Learning – Forward inferencing using Raspberry Pi equipped with Low cost Wide Angle Fish Eye Camera (US$27). Trained on desktop with 2x Nvidia 1080Ti. Radio Controlled Car (toy)

 

Video: Person following robot (RC Car), by Yeshas Thadimari, Lee Yi Yuan, and Yeo Jun Hao. Winner of  HacknRoll 2019. https://devpost.com/software/followmesempai-otc6mz

Running on Nvidia Jetson TX2 with ZED Camera
Also in youtube: https://youtu.be/ERCPDWpL_w0

Video: An omni-directional mobile base with powered caster wheels inspired by a swivel chair. Take a look at the wheel of your swivel chair: replace the two discs in one wheel with 2 independently driven hub motors. These 2 motors allow the rolling and steering of the wheel. (design by Haoyong YU)

Video: Omni-directional motion capability achieved using meccanum wheels - see our mobile robot manouever around cooridors. It's capable of a 200 kg payload

Video: A tracked mobile base that can traversed unstructured environment and climb up/down stairs. The body and flippers have independent tracks, which allow it to move using flippers, body-tracks or both.

Deep Learning and AI

Video: Recognition/Classification of Objects using Deep Neural Networks. Persons and cars (static and dynamic) are detected and tracked.

Video: Object classification and semantic mapping - in an office environment in NUS CREATE – Utown

Video: Robust Feature Matching in Varying Lighting Conditions (Learning Low Rank Images for Feature Matching, using our GAN (Generative Adverserial Network)

Video: Object Detection, Scene Classification and Semantic Mapping

Video: Deep Learning –  Simultaneous Detection and Tracking (speed of vehicles)

Video: Deep Learning – Tracking of Vehicles (ICRA 2018) – outputs velocity and distance,
(our algorithm also outputs  relative x, y, xdot, ydot of tracked vehicles)

Video: Deep Learning for Intention Prediction of Vehicles (ICRA 2018) – Will the vehicle(s) change lane?

Scene Recognition and Object Detection in A Unified Convolutional Neural Network

https://youtu.be/GcDCJbEKhhA

 

Point Cloud Processing for Tabletop Object Geometric Information Prediction

https://youtu.be/ggy-JIdHpY8

Autonomous Driving

Video: Safe Path Planning among Moving Obstacles (IROS 2019)

 

Video: End-to-End Deep Learning – Forward inferencing using Raspberry Pi equipped with Low cost Wide Angle Fish Eye Camera (US$27). Trained on desktop with 2x Nvidia 1080Ti. Radio Controlled Car (toy)

Video: Autonomous Wheelchair in Changi General Hospital, Singapore

Youtube or Video: Hop on and experience a driverless ride with SCOT!
SCOT, or Shared Computer Operated Transport, is Singapore’s first locally-developed driverless car, jointly developed by NUS Engineering and the Singapore-MIT Alliance for Research and Technology. SCOT is currently on trial within Singapore’s one-north business park.

Youtube or Video: Multi-class Mobility on Demand using Autonomous Vehicles

Youtube or Video: SMART has expanded its self-driving vehicle fleet to now include a personal mobility class with a mobility scooter. The scooter was demonstrated in the 2016 MIT Open House, with this video showing test runs from the National University of Singapore (NUS) University Town Plaza, MIT's Infinite Corridor, and footage from the MIT Open House exhibit at Massachusetts Avenue and Vassar Street parking lot.

Video: Personal Mobility Scooter avoiding pedestrians with a conservative safety margin

Youtube or Video: Mapping is done with a single 2D lidar that is tilted 15 degree below horizontal. You can see map adjusting itself when loop closures are encountered. We are able to localizate and map without GPS.

Youtube or Video: Generalized Predictive Planning for Autonomous Driving in Dynamic Environments. Self-driving vehicle planner stochastically generates coupled spatial paths and velocity profiles, collision checking over space-time around predicted obstacle trajectories. Generality shown by same algorithm applied for planning onboard three vehicle platforms (scooter, buggy, and road car), in varied environments (pedestrian and on-road).

Youtube or Video: V2V Communication is utilized to enable coordination for collision avoidance and for resolving deadlocks between multiple vehicles. We use a "coordination diagram" with localized optimization to avoid deadlocks.

Youtube or Video: Our first publc deployment of our autonomous buggies at the Japanese and Chinese Gardens in Singapore, October 2014.
news article here.

Youtube or Video: Tele-driving a car.

Youtube or Video: Why Autonomous Vehicles - for fun!

Mobile Robotics

See our in-house designed omnidirectional mobile robot following a person: video1, video2

Two-Stage Optimized Next-View Planning[MA1] 

https://youtu.be/kstuXMDhjnU

 

Obstacle-guided Informed Planning for Navigation in Cluttered

Environment

https://youtu.be/HnG5OsWoS0s

 

2-D Cluttered Corridor Navigation Instance

https://youtu.be/DdZn7FtC4g0

 

Active Path Clearing through Local Environment Reconguration

https://youtu.be/4CEnf8M15Xk

 

Target-driven Model Learning for Collision-aware Non-prehensile

Holonomic Object Pushing

https://youtu.be/KaGD4_PghlI

 

Integrated System for Navigation in Unstructured Environment

https://youtu.be/4KxXtATbT2Y

Marine Robotics – Autonomous Vehicles in Marine Environments

Bumbleebee Autonomous Underwater Vehicle (Version 3.5, 2nd Place during 2018 Robosub international competition, San Diego, USA
https://www.youtube.com/watch?v=B3KIOdm4y1k
Video

BumbeeBee Autonomous Surface Vehicle (V 2.0 Preview) (Version 2.0 – 1st place during RoboSub 2018 in Hawaii)
https://www.youtube.com/watch?v=6dRh1iljWT8
Video

 


 [MA1]